Research Methods¶

Methods for studying learning and evaluating education that are specific to it or especially useful in it. General social-science methods are not listed.

641 entries · 0 stable · 641 in review · 0 drafts

In Review¶

Jump to: A · B · C · D · E · F · G · H · I · J · K · L · M · N · O · P · Q · R · S · T · U · V · W · Z

A¶

A bottom-up taxonomy of seven student LLM usage categories organized into three support types - The article presents a refined bottom-up coding scheme of students' LLM use, derived iteratively from LLM-use reports in a research-oriented course and cross-labeled by student initiative.
A CAT-sensitive process for evaluating content alignment - The article outlines a process for evaluating content alignment that is sensitive to the particular ways computerized adaptive tests can diverge from their intended content domain.
A data-driven taxonomy organizing the meanings of concrete and abstract along a learner–representation–concept dimension - The article proposes a taxonomy of the semantic landscape of concreteness and abstraction in mathematics education, built from qualitative coding of a corpus of 368 papers.
A four-step adapted cognitive task analysis framework for clinical environments - The article develops a health care setting-adapted version of cognitive task analysis, a suite of methods for studying cognitive processes such as reasoning, problem-solving, and decision-making.
A four-step evaluative framework for judging educational practices and beliefs - The editorial describes a shared evaluation process modeled by every article in the special issue.
A four-type hierarchy of evidence for judging educational technology effectiveness - The guide presents an organizing framework of four evidence types that educators encounter when evaluating product effectiveness claims.
A framework and operationalization of study designs and statistical approaches for studying academic resilience in literacy - The special issue's theoretical backbone distinguishes promotive effects (factors benefiting all learners), protective effects (factors offsetting risk), and compensatory mechanisms.
A hierarchical taxonomy of novice difficulties with selection and iteration in code reading - The article introduces a fine-grained taxonomy categorizing novice programmers' difficulties with reading control flow constructs.
A priori taxonomy of three inference types defined by how interdependent information is distributed between collaborators - The article distinguishes three types of solution-relevant inferences, defined a priori by the experimental distribution of two interdependent pieces of information in a dyad: \"Collaborative inferences\" result from unshared, distributed information and \"can only be drawn collaboratively\"; individual
A taxonomy classifying authentic CSCL learning environments into simulation and hybrid environments with five descriptors - This conceptualization, from Hod and Sagy, categorizes and compares authentic learning scenarios.
A ten-strategy taxonomy of distractor-generation strategies grounded in learning-science literature and observed LLM reasoning - The article introduces a taxonomy of ten distractor-generation strategies, \"grounded in learning-science literature and empirical observations of LLMs’ reasoning traces\", spanning task understanding (INTER, CORR), candidate construction (LINK, ERR_DESC, ERR_SIM, INST), quality control and selection
A validity-oriented framework for comparing assessment latent structure across humans and LLMs - The article introduces a methodological framework that treats similarity of latent factor structure as a necessary, though not sufficient, condition for transferring score interpretations from humans to LLMs.
Active/passive local/global coding framework for LLM offloading prompts - A four-way coding scheme applied to each of 564 learner prompts, capturing whether the learner \"remained somewhat active or was passive during a comprehension activity, and if the comprehension within a prompt was targetting local or global level\".
Actor-perspective by setting taxonomy for classroom eye-tracking research - The article organizes eye-tracking research in education along two axes: the actor perspective studied (teacher vs.
Adapted conjecture map framework for learning scientists and computer scientists - An adaptation of Sandoval's conjecture mapping tool, created by Chang and Roschelle, that supports interdisciplinary design-based research teams in connecting learning theory to technical design.
Adapted IRC framework for novice students' interdisciplinary reasoning - The study adapts Shen et al.'s (2015) interdisciplinary reasoning and communication (IRC) framework for novice high-school students rather than experts.
Adapted Personalized Learning Continuum Framework (PLCF) - The article adapts the Personalized Learning Continuum Framework of Lokey-Vega and Stephens (2019) as an analytical lens for classifying AI-enabled personalized learning models along the proportion of Academic Learning Time spent on automated versus student-centered pedagogies and the distribution o
Adaptive G-UKT - Adaptive G-UKT is a knowledge tracing framework that models each learner's evolving latent state as a Gaussian distribution rather than a point vector, tracking \"semantic activation levels and estimation confidence through diagonal covariance\".
Agent-based educational science (AES) - Agent-based educational science (AES) is a research agenda in which educational theories are formalized as interacting agents and environments, enabling in silico experimentation on developmental processes.
Agent-based simulation methodology for decomposing teacher quality gap sources - The article introduces \"a novel simulation methodology\" that models teacher workforce dynamics to attribute teacher quality gaps to their underlying processes.
Agent-characteristic moderator taxonomy: appearance, gender, nonverbal communication, motion, and voice - The article organizes its meta-analytic tests around a set of agent characteristics that \"may moderate their instructional effectiveness, including appearance, gender, nonverbal communication, motion, and voice.\" This five-branch taxonomy structures which agent design features were hypothesized (by
Ambulatory assessment with intensive longitudinal data - Ambulatory assessment is a measurement approach in which intensive longitudinal data are captured in real-life settings with repeated within-subject assessments of physical activity (e.g., via accelerometry) and cognitive performance (e.g., via a smartphone app) at high measurement density, such as
AMOVA (Accumulative Manifold Validation Analysis) - AMOVA (Accumulative Manifold Validation Analysis) is a specialized statistical methodology for testing the internal and external validity of uniquely designed psychometric instruments.
Approach–context–pedagogy coupling framework for game-based and gamified robotics education - The review's organizing framework holds that instructional approach (GBL vs.
Aptitude-treatment interaction (ATI) framework - The article organizes its moderation analysis around the aptitude-treatment interaction (ATI) paradigm, which \"states that the effectiveness of an intervention can be conceptualized as a result of the complex interplay between the distinct features of the intervention and students' ability and willi
Architecturally enforced novice-like student simulation - The article's central design recommendation is that realistic synthetic student behavior cannot be elicited from preference-tuned LLMs by persona prompting or context injection alone; it requires architectural enforcement.
Arts-based research (ABR) - The review proposes a shared definition to address the \"confusing plethora of terms\" in the field: \"The use of artistically inspired methods by researchers and participants in a collaborative research environment where members of the community are actively involved either in creating art in the sear
ASIMeC analysis scheme for coding message characteristics and role adoption - ASIMeC is a content analysis scheme developed by the authors to identify message characteristics along five dimensions: moderating, summarising, adding new discussion points, adding theory, and adding external information.
Assessment metadata as naturally occurring direct SEL measurement - The article proposes that assessment metadata — data about how students behave on a test or survey, including disengagement indicators — can function as naturally occurring, direct measures of social-emotional competencies, offering an alternative to deliberately administered SEL surveys.
ATC21S collaborative problem-solving framework - The study applies the ATC21S CPS framework, in which \"CPS is defined as a set of elements consisting of 'social' and 'cognitive' as the first-order components\".
Attainment-based evaluation of charter schools as a complement to achievement-test-based studies - The report frames its contribution as moving beyond the standard outcome measure in charter school research.
Augmented cross-recurrence graphs for distinguishing balanced and unbalanced participation - A new visualization developed by the authors for dual eye-tracking data.
Automated analysis of argumentative discourse for diagnostic learning analytics - The article identifies automated analysis techniques for detecting learners' internal argumentative scripts during CABLe.
Automated computer-vision analysis of gaze and proxemics from classroom video - The study adapted the Sharingan gaze-estimation framework (Tafasca et al., 2024), a pretrained transformer-based model reported with AUC 0.944, to derive joint attention and mutual gaze frame by frame from standard 2D video.

B¶

Bakhtinian dialogic framework of four interconnected themes — voices, expanding the dialogue, inter-animation, and relations - The article operationalizes dialogue through four Bakhtin-inspired themes used both to train teachers and to analyze noticing utterances.
Baseline covariate adjustment in education randomized trials - The paper recommends that studies using student achievement outcomes generally aim to collect and use baseline data, because analyses using more than one wave of data typically have much greater statistical power than posttest-only analyses.
Bayesian Networks for student modelling, assessment, and collaborative learning theory - Bayesian Networks are presented as both a specification language and a computational mechanism with practical value in designing group and learning support systems: they have \"been frequently used to build student models (Millán et al., 2010) and for assessment (Almond et al., 2015)\".
Behavioral choice measure of maintained situational interest - The article operationalizes maintained situational interest not by verbal self-report but behaviorally: participants decided which of six described seminar topics (Experiment 1) or which of six take-home texts (Experiment 2) they wanted to engage with, and the count of yes-responses was the measure.
Behavioral coding methodology for identifying epistemological framing in group video - A real-time observational coding protocol in which coders watch video of small groups without a transcript and note the exact time of changes in vocal register, affect, grammar, gesture production, and body language.
Behavioral telemetry logging with observable, unambiguous, contextualized events - The chapter holds that telemetry \"should have the following properties: (a) is a description of behavior and not an inference about why behavior occurs, (b) is unambiguous,\" and (c) contains sufficient context to link the element to a specific player at a specific point.
BERT strategy-embedding and successful-strategy prediction from MATHia logs - A BERT-based model the authors train to learn embeddings for math strategies from MATHia student log data, together with a prediction task that distinguishes between successful and unsuccessful strategy use.
Between-student testing and response-curve method - The between-student method, first used in physics by Sayre and Heckler, gives short conceptual quizzes to different sub-groups of the population in successive weeks, avoiding test-retest effects.
Bidirectional validity approach - The bidirectional validity approach examines a test's fairness by working backward from the criterion: examinees are matched on the outcome the test is supposed to predict (here, type of math course taken and grade received in first-year college math), and their test scores are compared.
BIT Coding method for Newton's Third Law responses - The BIT Coding method is a new method of analysis created by the authors \"to quickly identify students' understanding of Newton's Third Law questions.\" It was employed alongside the Force Concept Inventory gain analysis in this study of 222 introductory physics students, complementing the overall an
Bloom-aligned framework for measuring educational control in LLMs - The framework operationalizes educational control — \"the ability to preserve a task's instructional intent while shifting its cognitive demand toward specified learning objectives\" — using revised Bloom's Taxonomy as an ordinal scale from Remember (1) to Create (6).
Bloom-driven cognitive-leap evaluation framework with CogShift, category consistency, and knowledge coverage metrics - A Bloom's-Taxonomy-driven framework for evaluating LLM-generated educational questions that pivots \"from a knowledge-point perspective to a cognitive-leap paradigm\".
Body of work approach to standard setting for alternate assessments - The body of work approach is a standard setting methodology in which \"Reviewers examine all of the data for a student and use this information to place the student in one of the overall performance levels.\" Standard setters receive papers demonstrating the complete range of possible scores from low
Bootstrapping adjustment (BA) method for item parameter bias reduction - BA is one of two item parameter adjustment methods proposed in the article, reducing biases in both item difficulty and item discrimination parameters using statistical bootstrapping.
Brennan and Resnick's three-dimensional CT framework (concepts, practices, perspectives) - The article adopts Brennan and Resnick (2012)'s framework as its theoretical scaffold for understanding CT learning and assessment.
Brier-score decomposition into calibration, resolution, and knowledge as metrics of detection metacognition - The article uses probability judgments to reveal metacognitive states of mind, scoring them with the Brier score, which \"can be decomposed into three components, calibration (or reliability), resolution, and knowledge (or uncertainty).\" Calibration measures whether confidence judgments correspond to

C¶

C.Ind.Le Coding Scheme - The study's measurement approach applies the C.Ind.Le Coding Scheme to individual interviews of six-year-olds during an ecological and challenging graphic task.
CAPERS (Community and Participatory Evidence and Research Synthesis) - CAPERS (Community and Participatory Evidence and Research Synthesis) is a mixed-methods framework that engages community partners prospectively in research synthesis rather than post hoc.
Categorization schema for classroom factors: density dependency and psychosocial vs. environmental - The framework operationalizes classroom factors along two dimensions: density dependency and type.
Causal mediation analysis of teacher practice in education randomized trials - The report defines a mediator as \"an intermediate outcome that is measured after random assignment and that can be affected by the treatment.\" Its Figure 2.1 path model posits that intervention services improve teacher knowledge, which improves teacher practices, which improve student test scores, w
Causal-chain analysis of built-environment effects - The brief describes \"causal chains\" as a method environmental health researchers developed to evidence how building features affect people where direct causal proof is impractical.
Cautious interpretation of within-domain effects in the classic I/E model - The article recommends that researchers avoid interpreting within-domain effects in the classic I/E model as exclusively reflecting social comparisons.
Centrality measures as operationalization of the participatory and social dimensions of CSCL - The article organizes SNA-based CSCL measurement around two dimensions.
Chain of assumptions/inferences perspective within an argument-based approach to validity - The article works within an argument-based approach to validity, in which the chain of assumptions and inferences connecting test scores to interpretations and uses is made explicit.
CID framework for the cognitive analysis of composite instructional designs - The CID framework (a framework for the cognitive analysis of composite instructional designs) explains how learning in one phase affects subsequent phases by focusing on the learner's knowledge state between phases.
Class composition effects framework - The article organizes its inquiry around composition effects, which \"occur when individual student characteristics (e.g., individual academic achievement) are combined into a class-level composite score (e.g., mean academic achievement) and that score is related to individual-level outcomes (e.g., a
Classical Test Theory - CTT (True Score Theory) describes an observed score as composed of a true score and random error, estimating true scores from observed scores.
Classical test theory decision consistency estimation methods - The article organizes single-administration decision consistency estimation around three CTT methods: Livingston and Lewis (LL-DC), which assumes a beta-binomial distribution and uses effective test length; Peng-Subkoviak (PS-DC), a normal approximation extending Huynh's procedure; and Wolkowitz (W-
Classification framework distinguishing global quality, global process quality, quality pedagogical interactions, and quality interactions related to curriculum content - The review's organising taxonomy for grouping interaction quality characteristics.
Co-referencing typology and coding scheme for analyzing collaborative interaction around external representations - The Evans, Feenstra, Ryon and McNeill article provides analytic tools for collaborative interactions around external representations, honing in on \"co-referencing\": multiple people referencing the same thing through a deictic word, a pointing gesture, or any verbal, physical, or virtual action refer
CoDiL framework: taxonomy of learning activities for digitally enriched educational settings - The CoDiL framework enumerates learning activities that digital instructional features may initiate and stimulate, each with an external behavioral side (observable student-tool interactions) and an internal cognitive side.
Coding framework of metacognition intervention components and instructional methods - The article develops a fine-grained coding framework for metacognition interventions, distinguishing intervention focus (primary vs.
Coding scheme for observable Direction, Integration, and Evaluation in student–AI interaction traces - An exploratory multilabel coding scheme applied to student turns in AI-mediated conversations, with the student turn as coding unit interpreted in relation to its local sequence and the whole conversation.
Coding scheme of young learners’ metacognitive SRL activities with domain-general and task-specific sublevel categories - A think-aloud coding scheme for metacognitive SRL activities in lower secondary students, developed from theory and data using Bannert’s (2007) and van der Stel and Veenman’s (2014) schemes as a basis.
Cognitive and social-emotional perspective-taking coding scheme for dyadic AUT dialogues - A novel observational coding scheme designed by the authors, based on the PAT theory and inspirations from Chiu (2008), applied utterance-by-utterance to videotaped and transcribed AUT sessions.
Cognitive Diagnostic Profiling (CDP) - CDP is a zero-shot framework that draws on cognitive diagnostic models \"to represent each simulated examinee as a binary attribute-mastery profile and prompts an LLM to generate item responses consistent with that profile.\" It proceeds in six steps: construct decomposition into binary attributes, ma
Cognitive episodes: an eight-way taxonomy of functional problem-solving states grounded in Schoenfeld's Episode Theory - Cognitive episodes organize LLM reasoning traces into temporally ordered functional problem-solving states.
Coherence analysis framework and its six activity-trace measures - Coherence analysis (CA) is the paper's approach to analyzing learners' problem-solving processes in OELEs by \"combining information gleaned from sequences of student actions to produce measures of action coherence.\" CA assumes that when students encounter information that could improve their solutio
Cohesion-graph model for automated collaboration assessment - The article's first proposed model represents collaboration through \"a cohesion graph\" over the conversation.
Cohort-based graduation and dropout rate calculation - The report calculates graduation and dropout rates by following student cohorts over multiple years using two cohort types.
Collaboration Literacy Analytics Framework (CLAF) - The CLAF is a conceptual framework 'that integrates evaluation metrics, indicator interdependencies, and feedback mechanisms.' It 'captures the dynamic relationship between process and outcome measures, connects individual- and group-level indicators, and incorporates review mechanisms as student-dr
College-readiness predictor selection using decision trees, random forests, lasso, and factor analysis - This framework uses data reduction to select, from thousands of measures, the variables that maximize classification accuracy for college enrollment and persistence.
Combined fine-grained video coding and process mining for sequential collaborative-learning analysis - The article introduces a methodological contribution that combines fine-grained video analysis with process mining to study sequential patterns linking challenges and collaborative regulatory activities.
Combined regression discontinuity and interrupted time-series design with instrumental variables - When a threshold-based policy changes conditions for non-targeted students too, a standard regression discontinuity cannot estimate the policy effect, only the enrollment effect.
Combined tSSNA and ONA method - This analytic method combines temporal socio-semantic network analysis (tSSNA) and ordered network analysis (ONA) to analyze collaborative knowledge creation.
Common acceptability thresholds for psychometric evidence in educational measurement - The review assembles a framework of psychometric indicators and their common acceptability thresholds in educational research, covering content validity, structural validity, internal consistency, reliability, and criterion validity.
Comparative framework for longitudinal models differing in change processes, temporal structure, and measurement design - Beyond responding to the commentary, the authors \"present a generalizable framework for comparing longitudinal models that differ in their assumptions about change processes, temporal structure, and measurement design.\" The framework organizes model comparison around three axes: what change process
Comparative guide table of AI model families for cognitive-model selection and validity checks - The review provides a summary table (Table 1) comparing the four major AI model families as cognitive models for educational psychology.
CoMPAS framework: six forms modeling individual participation in collaborative learning - The CoMPAS framework models individual participation in collaborative learning processes across six components: Cognitive interactions, Coordinative interactions, Metacognitive interactions, Passive participation, solo Active participation, and Socio-emotional interactions.
Component-based instructional design framework for analyzing ICT-supported writing interventions - The article adopts a component-based instructional design perspective, coding each intervention on four design aspects: Merrill's instructional principles (problem centered, activation, demonstration, application, integration), the instructional activities that operationalize them, the types of feed
Compound Polynomial growth model - The CP model, developed by Thum and Hauser (2015), is a flexible functional form for describing growth trends with marked seasonal patterns.
Comprehensive checklist of reporting criteria for intensive longitudinal studies - The review offers \"a comprehensive checklist of reporting criteria for intensive longitudinal methods\", integrating the APA Journal Article Reporting Standards, the Trull and Ebner-Priemer (2020) guidelines, and the review's own findings.
Computational formalisation of CSCL theories using Bayesian models - The authors recommend striving for a unified language, such as a formalisation strategy, in which different CSCL theories can be expressed, proposing the Bayesian framework as a formal language within reach for CSCL researchers.
CoMSA three-parameter logistic learning-curve model - The CoMSA conceptualizes educational differences through three logistic curve parameters.
Confidence-aware selective automated scoring framework - The framework adapts a pretrained ViT with LoRA and derives a response-level confidence score from the stability of predictions under semantic-preserving test-time perturbations (crops, rotations).
Consensus calibration - Consensus calibration treats recalibration as a divide-and-conquer Bayesian computation: each time period is calibrated independently and the period posteriors are combined without revisiting earlier data.
Construct evaluation framework - A framework for evaluating the characteristics of psychological constructs prior to automated coding, grounded in educational psychology measurement practices and validity theories (Cronbach & Meehl, 1955; Kane, 2013; Messick, 1995).
Constructivist three-role taxonomy of learner-generated questions - The article organises learner-generated questions through a constructivist framework drawing on Piaget and Vygotsky's theories, in which student inquiries are categorised into three roles: \"knowledge transmitter, facilitator and co-learner\".
Contemporaneous (lag0) sensitivity analysis for selecting panel-design time intervals - Because effect sizes vary with the interval between measurement waves and there is no optimal interval, the article proposes testing lag0 paths within each wave as a practical check on design.
Contextual similarity scoring model for automated essay grading - The article's scoring framework matches a student's response to an examiner-supplied model answer using semantic contextual search similarity, returning a similarity index of 0–1 that is multiplied by the item's allotted marks.
Contingent shift framework for measuring scaffolding contingency in classroom interaction - The contingent shift framework, based on Wood et al.
Continuous Conjunctive Model - The CCM is an extension of the NIDA cognitive diagnostic model that \"utilizes continuous variables instead of binary latent variables to represent attributes, thus it offers a more generalized diagnosis\" (Hong et al., 2015).
Continuum for classifying instructional quality frameworks: generic, subject-specific, and hybrid - The article organizes frameworks for conceptualizing instructional quality along a continuum adapted from Charalambous and Praetorius (2018), in which \"Senden et al.
Contrapuntal reading - Building on Said (1993), the article defines contrapuntal reading as drawing out \"the logics of and resistances to empire and imperialism that are silent or marginally present\" in designs and theories.
Controlled asymptotic practice-curve method for transfer experiments - The article argues that earlier transfer studies were unconvincing because training was too short and controls were missing.
Coordination-based evolutionary game-theoretic framework of student AI use in assessment - The framework models student generative-AI use in assessment as a two-player symmetric coordination game in a well-mixed finite population, in which the institution is not a strategic player but shapes payoffs through assessment design.
CORDTRA diagrams for visually representing the chronology of tool use and discourse in collaborative learning analysis - CORDTRA stands for Chronologically-ordered Representations of Discourse and Tool-Related Activity.
Cosine similarity framework - The article introduces an angular-statistics framework in which item and test difficulty and discriminating values are obtained from the cosine of the angle between the observed score vector and the maximum possible score vector, involving inner products and vector lengths.
Countable versus dispositional prompt-rule audit - The paper's organizing framework splits system-prompt instructions into two kinds.
CourseGraph course-overlap detection methodology - CourseGraph is a four-step methodology for automatically assessing whether an external course overlaps with courses in a home curriculum.
CPIG: the creative psychometric item generator framework - CPIG is an iterative LLM-based framework for generating and validating creativity test items for the creative problem-solving (CPS) task.
Creswell and Plano Clark's three core mixed methods designs (convergent, explanatory sequential, exploratory sequential) - A taxonomy of three basic mixed methods research designs in which researchers combine quantitative and qualitative approaches in a single study.
Criteria for identifying spatial factors - The article establishes inclusion and exclusion criteria for calling a factor a spatial factor, because factor names do not always align across studies and knowledge-laden factors have been misclassified.
Critical interpretive synthesis underpinned by Kirkpatrick's return on investment model - This is the review's methodological framework: a \"Critical interpretive synthesis underpinned by Kirkpatrick's return on investment model\" applied to identify features of effective dementia educational programs.
Cross-theoretical integration of four motivational theories of teacher influence on engagement - The article synthesises four theories of how teachers improve student engagement: self-determination theory, achievement goal theory, growth mindset theory, and transformational leadership theory.
Curricular analytics: graph-based measures of degree-program complexity (blocking, delay, centrality, structural complexity) - Curricular analytics is the study of academic program design and layout, in which a degree plan is translated into a directed acyclic graph of courses and prerequisites.
Curriculum analytics using cognitive diagnostic modeling and hierarchical learner-profile clustering - A two-step curriculum analytics approach that maps existing assessments to professional standards and applies a generalized cognitive diagnostic model to assessment grades to produce attainment probabilities (0–1) per standard, then uses hierarchical agglomerative clustering to group learners into i
Curriculum-grounded ontology framework for building educational QA benchmarks - The article proposes an end-to-end framework in which a national curriculum textbook is ingested via OCR and semantic chunking, formalized against a predefined domain ontology, and used to construct a provenance-traced QA benchmark with category-level evaluation.
Curriculum-Instruction-Assessment triad taxonomy of AI educational purposes in VET - Building on Wang et al.

D¶

Data sessions and multiple experienced analysts for intersubjective validation - Stahl proposes data sessions—involving multiple analysts reviewing the same data—as a mechanism for guarding against bias and idiosyncrasy in EM case analysis.
DeCuir-Gunby and Schutz's five purposes for considering race in educational psychology research - The review adopts DeCuir-Gunby and Schutz's (2014) classification as a common language for analyzing each article's approach to race.
Deriving observational self-regulation coding items from adult-reported measures - The review recommends grounding observational tool development in analysis of adult-reported measures because such tools emphasize overt behaviors across diverse daily contexts and informants.
Design Science Research (DSR) for CSCL - The editors propose importing Design Science Research from Information Systems into CSCL.
Design-Based Implementation Research (DBIR) - DBIR is an emerging research-and-development approach the article positions as the process for co-developing its differentiated instruction practice framework.
Design-based RCT analysis with grouped administrative data - The article recommends analyzing RCTs with grouped administrative data \"to help improve data access,\" using design-based estimators formed as group-level averages.
Design-parameter matching in a priori power analysis - This principle holds that a priori power analyses should strive for an ideal match between the selected design parameters and the planned study's characteristics: the clustering structure (single-, two-, or three-level), the target outcome domain, the assessment method, and the target age group.
Design-thinking methodology for improving and scaling psychological interventions - The article formalizes a methodology for preparing psychological interventions to be scaled, combining two design traditions.
Detailed methodological reporting for social network analysis in learning analytics - The article recommends that researchers reporting on social network analysis in learning analytics document their operationalization choices, because measurement robustness underpins reproducibility.
Dialogical self theory (DST) I-positions as a CSCL analytical framework - DST holds that, as the article states, \"the self is populated by other inner selves that collectively derive from the world and have their own voice, motives, opinions, and knowledge that become present in what are known as I-positions via dialogue in response to the social situation\".
Dialogical voice-synergy model of collaboration - The article's second model assesses collaboration \"from a dialogical perspective\" as the participation of different speakers through their voices.
Dialogue Act Coding (DAC) system - The DAC system identifies \"dialogue acts\", the communicative function of each utterance typed during online collaboration, building on the Verbal Observation System and discourse-analysis research.
Difference I/E model - The difference I/E model, proposed by Wolff (2021), is a reparameterization of the classic I/E model in which \"self-concept in one domain is regressed on achievement in the same domain and on the difference between achievement in the same domain and achievement in the other domain.\" This operational
Diffusion-based centrality framework for measuring uptake in CSCL discourse - A framework representing threaded CSCL discussion as a directed post-reply network in which centrality measures operationalize the uptake of ideas and stimulation of further discussion.
Digital learning platform process-data analysis - The report urges researchers to use the robust log data DLPs generate, such as which tasks students begin, time spent, and detailed progress.
Digital shadowing as a framework for recording LLM behaviour under controlled conditions - Digital shadowing is the article's organising methodological framework: unlike digital twins, shadows record system behaviour without feedback loops.
DINA and DINO cognitive diagnostic models - The article identifies the DINA model and the DINO model as \"two typical theoretical diagnostic classification models\".
Direct and soft disagreement taxonomy for coding peer critique talk - The study adapts Pomerantz's strong/weak disagreement distinction into a two-category coding scheme for dissenting turns in peer critique.
Discrimination-index Q-matrix validation for cognitive diagnostic models - The article supplies the theoretical basis for using its discrimination index as the foundation of a Q-matrix validation method.
Divergent Collaborative Learning Mechanisms (DCLM) framework - The editorial describes Tissenbaum, Berland, and Lyons' framework \"for recognizing and coding collaboration and divergent learning — Divergent Collaborative Learning Mechanisms (DCLM)\", used to analyze museum visitors' interactions with an interactive digital tabletop exhibit.
DMO-GRU hybrid student performance prediction method - The DMO-GRU framework combines the Dwarf Mongoose Optimization algorithm with a Gated Recurrent Unit neural network.
Dose–response framework with dosage defined as cumulative intensity - The article applies the dose–response framework, borrowed from medicine and pharmacology, to academic intervention: the dose–response curve represents the relation between an intervention's intensity and its effects.
DReaM framework: nine sub-categories of uncertainty articulations integrating statistical, scientific and nature of science uncertainty - DReaM (Deterministic Relativistic and Middle ground) is an integrative framework offered by the authors that organizes learners' uncertainty articulations into nine sub-categories spanning three reasoning domains.
Dual-assessment framework for evaluating LLM human simulations via joint self-report and observer-rated persona measurement - A methodology adapted from clinical psychology that jointly measures self-reported characteristics and observer-rated behavioral expressions of an LLM persona.
Durkin's eight-category scheme for classifying teachers' comprehension-related behavior - The article develops a coding scheme distinguishing comprehension instruction—\"Teacher does/says something to help children understand or work out the meaning of more than a single, isolated word\"—from seven related categories: application, assignment, help with assignment, review of instruction, pr
Dynamic structural equation modelling for intra-individual emotion trajectories - The study advocates measuring emotions repeatedly after every session and analysing them with dynamic structural equation modelling, a process-oriented, intra-individual approach \"only recently applied in educational research\".
Dynamical-systems analysis of intelligent tutoring system process data - The article proposes dynamical systems theory and associated analysis techniques as tools for examining behavioral patterns and variations within ITSs.

E¶

Early-warning indicator construction - A Consortium report cited in the paper outlines five characteristics of effective indicators: predictiveness, usability and clarity, real time/right time availability, direct causal linkage to outcomes, and malleability.
Ebbinghaus savings-paradigm variation for detecting early incidental vocabulary gains - This methodology exposes learners to novel words incidentally, then asks them to learn the words explicitly, with half of the explicit-phase words previously encountered and half completely new.
EduBehaviors - EduBehaviors is an interpretable annotation framework in which researchers write explicit binary assertions about conversational behavior; LLMs annotate each assertion as true or false, and a rule or trained statistical model combines these values into a construct label.
Educational characteristics framework for analyzing democracy-teaching interventions - Beyond teaching-method types, the review analyzes interventions through three educational characteristics.
EEG-based continuous online cognitive-load measurement - Antonenko et al.
Effort analysis - Effort analysis is a method for conducting \"effort-related individual score validation\" of achievement test data.
Effort-moderated (E-M) scoring model - The effort-moderated (E-M) model, attributed to Wise and DeMars (2006), is a scoring approach in which \"rapid guesses are identified and excluded during scoring.\" It is \"intended to estimate the score a disengaged test taker would have received had they been fully engaged.\" The presentation evaluate
Eight validity criteria for AI systems as cognitive models of learning - The review's central contribution is an evaluative framework of eight criteria that an AI system must satisfy before its cognitive-model status is granted rather than assumed: theoretical grounding, construct validity, mechanistic transparency, alignment with human learning trajectories, error-signa
Eight-stage synthesis process with participatory integration actions at each stage - The article maps community integration onto eight synthesis stages: planning and protocol development, literature search, study identification, coding and data extraction, data analysis, results interpretation, reporting, and dissemination.
Eight-theme analytical framework for GenAI in higher-education group work - The review constructs eight recurring themes via reflexive thematic analysis of 18 studies: GenAI reconfigures knowledge development, reshapes higher-order thinking, mediates communication, transforms team interaction, improves administrative efficiency, provides personalised feedback, and raises da
Elaborative style: an utterance-level coding framework for adult–child reminiscing - Elaborative style is a dimension of adults' language assessed in reminiscing research, classified on a continuum from high to low.
Eleven-component taxonomy of teacher feedback during oral reading - The review codes the feedback practices in 24 K–5 oral-reading studies into a taxonomy of feedback content.
Embodied learning taxonomy of sensorimotor engagement, gesture congruency, and immersion - The introduction reports, citing Johnson-Glenberg et al.
Emergent LLM coding plus embedding-based consolidation pipeline for large-scale student-discourse taxonomies - A two-step methodological framework, framed as the LLM analog of open coding in grounded theory, for converting large student-tutor transcript corpora into interpretable discourse categories.
Emote-aloud procedure for on-line identification of affective states during learning - The emote-aloud procedure is a modification of the cognitive-science think-aloud method in which participants state aloud the affective state they are feeling while working on a learning task.
Empirical Bayes shrinkage for teacher value-added estimates - Empirical Bayes shrinkage is described as \"an approach commonly used in implementing teacher accountability systems\" that adjusts value-added estimates.
Engagement forecasting as a supervised week-ahead prediction task in ITS - Engagement forecasting is framed as a supervised prediction task estimating a learner's next-week engagement from prior ITS interaction traces.
Engagement-based evaluation framework for AI tutors (RelScore and SuccScore) - The framework extends AI tutor evaluation with a behavioral dimension grounded in student interaction data, complementing pedagogical assessment.
Engagement-probe ratings as a pragmatic single-item operationalization of momentary attentional engagement - The article frames its single-item probe (\"How difficult was it to pay attention during the last minute of the lecture?\") as a temporally localized reference signal targeting perceived difficulty sustaining attention, explicitly not equivalent to engagement as a whole.
Epi2Diff framework: episode-induced process representations combined with semantic embeddings for item difficulty prediction - Epi2Diff (Episode to Difficulty) converts model-generated reasoning traces into sentence-level episode sequences and extracts compact features capturing reasoning effort.
Epistemic network analysis (ENA) - Epistemic network analysis (ENA) is a non-parametric analytic method developed for epistemic game data; the article focuses on its social-network based variant.
Epistemological Move Analysis (EMA) - Epistemological Move Analysis is an analytical approach that analyzes a teacher's role in students' learning process, used in this study to examine how the teacher directed children's learning.
Equal-difficulty word-list methodology for classroom spelling experiments - The article describes a methodology for constructing and validating two lists of spelling words of equal difficulty for a given class.
Equipercentile linking with conditional growth norms for deriving assessment cut scores - The article's methodology combines the equipercentile linking method with NWEA's conditional growth norms.
Equity-oriented research methodology - Drawing on the National Academies of Sciences 2022 report, the article frames equity-oriented research methodologically in two ways: \"1) to look at differences between groups and 2) to design conditions that can enhance the performance of historically and systemically excluded groups.\" The report al
Error-type profiling - Error-type profiling is the analytic lens the article introduces to judge fairness interventions by how they redistribute false positives and false negatives across groups, rather than only by whether fairness metrics converge.
ESSA four-tier evidence hierarchy for intervention selection - The guide explains that \"ESSA defines four levels of evidence\" as elaborated in the U.S.
Event-centered analysis of CSCL processes - Peter Reimann proposes an event-centered approach as an alternative to conventional variable-centered methodologies for analyzing processes in CSCL settings.
Event-centered phase segmentation for multimodal simulation analysis - A reusable analytic pattern in which a simulation scenario is divided into three scripted segments so behaviors, regulation strategies and physiological responses can be compared across critical points.
Evidence-Centered Design - ECD is a principled assessment design framework that structures the rationales, choices, and consequences of assessment development.
Evidence-grounded multimodal lecture-to-KG pipeline - An end-to-end pipeline combining \"timestamped automatic speech recognition (ASR), high-recall anchor selection, OCR, grounded vision-language extraction, validation, canonicalization, KG construction, and GraphRAG-style retrieval.\" Its design constraint is that \"a concept or relationship is promoted
Evidence-grounding extraction schema for auditable educational knowledge graphs - The paper's organizing framework is an extraction schema in which every extracted item retains \"evidence, confidence, modality, lecture, timestamp, frame, and anchor identifiers.\" A concept records name, definition, evidence quotation, source modality, and confidence; a relationship records source,
EVOL: simulator-guided demonstration learning for deployment-free learning path recommendation - EVOL adapts the robotics sim-to-real demonstration recipe to educational recommendation: a DKT-based knowledge evolution simulator (KES) both synthesizes per-learner expert paths via evolutionary search and hosts RL fine-tuning, while the deployed actor plans from the initial mastery alone.
Executive LLM adaptive assessment method - The Executive LLM is a single LLM that generates responses for all AI teammates, has access to the pedagogical rubrics, and is prompted to \"create situations that will allow the subject to demonstrate their skill, so that their skill level can be quantitatively inferred\".
Expert-panel interpretation of compiled educational evidence - The article describes an evidence-review approach in which a research organization compiles a large body of studies and convenes expert panels to interpret the data.
Explainable-by-design framework coupling dual-lens hierarchical attention, trust-gated inference, and on-the-spot explanation generation - The framework is an LLM-based teacher assessment architecture in which trustworthiness and explainability are built into inference rather than appended afterward.
Explanatory Item Response Theory (EIRT) - EIRT is described as \"a special name given to models of multilevel, categorical responses, in which latent estimates are generated for both item and person parameters and explanatory variables can be modeled as predictors of responses across levels of analysis\".
Explicit assumption specification when comparing alternative longitudinal models - This principle holds that debates over alternative longitudinal models should distinguish defensible modeling choices from specification errors that depart from SEM logic.
Explicit vs. implicit interventions as a two-branch taxonomy of school-based creativity support - The review organizes school-based creativity support into two branches.
Exponential Learning with Sigmoidal Decay (ELSD) model of syntax acquisition - The ELSD model represents syntax acquisition as \"a simple exponential learning process\" in which grammatical proficiency grows with experience, with an experience discount factor modeled separately for simultaneous bilinguals, immigrants, and non-immersion learners.
Exponential random partition model for student team formation - The article applies an exponential random partition model to the formation of student teams in higher education.
Extended Skill Development Model (ESDM) - The ESDM is the authors' alternative to the reciprocal effects model of academic self-concept and achievement.
Extended Strobel taxonomy of five student–GenAI entanglement types - The paper extends Strobel et al.'s (2024) taxonomy of GenAI applications — originally derived from 100 commercial and professional applications — into a classification lens for educational settings.
Extended Taxonomy Design Process (ETDP) - The ETDP, described by Kundisch et al.
Eye Movement Sequence (EMS) model for identifying reading sequences from fixations - The EMS model is an algorithm the authors developed for an earlier study to identify reading sequences from eye-tracking fixation data over look zones.
Eye-tracking paradigm for studying peer feedback processing and use - A laboratory paradigm in which 59 higher education students read the feedback of two peers on a fictional essay and revise it while their gaze behaviour is recorded.
Eye-tracking research paradigm for studying professional vision and teacher expertise - The article characterizes an emerging research paradigm in which eye-tracking methodologies are used to study professional vision and teacher expertise.

F¶

Fake Question Paradigm - The Fake Question Paradigm is an objective behavioral measure of cheating in real exams.
FATE framework - The article organizes the evaluation of multimodal learning analytics around four dimensions synthesized from prior FATE research in AI and learning analytics.
Fight-pair comparison - The article's methodological framework is a within-pair comparison design: rather than comparing discipline outcomes across schools or students generally, it compares the punishments given to the two participants in the same fight, one Black or poor and one White or nonpoor.
Five control criteria framework for judging degree of control in media comparison studies - The review proposes that a media comparison is adequately controlled when five criteria are met: matched activities, matched non-IVR activities, matched practice with the dependent measure, matched time spent learning (within 20% difference), and matched content.
Five mixed-methods integration functions organize CAPERS: convergence, complementarity, development, expansion, and sampling - CAPERS organizes the integration of participatory and synthesis methods through \"five mixed-methods functions (Palinkas et al.
Five-class taxonomy of knowledge sources for synthesizing STEM PLC knowledge - The synthesis organizes its evidence into five named classes of knowledge: Type 1, elaborated empirical studies since 1995 mostly in peer-reviewed journals; Type 2, empirical studies reported without method detail and non-empirical work such as theory building; Type 3, published expert knowledge and
Five-component framework of quality formative feedback on writing - The study operationalizes quality formative feedback on writing using five deductively coded components: the degree to which feedback \"was criteria-based, provided clear directions for improvement, was accurate, prioritized essential features, and used a supportive tone.\" Each component is scored 1-
Five-component taxonomy of collaboration scripts - The article proposes that collaboration scripts can be analyzed into five components, which serve as a basis for comparing different script approaches.
Five-hypothesis validation process for online peer assessment - The article proposes a validation procedure in which peer assessment may only be judged valid after five prior conditions are confirmed: teacher reliability, correct test items, internal consistency of the test, reliability of the observation and marking tools, and clear, coherent marking instructio
Five-level diagnostic taxonomy of human–AI teaming defined by observable affordances - The chapter proposes a diagnostic taxonomy of human–AI teaming with five levels: transactional, situational, operational, praxical, and synergistic.
Five-level scaffolding scale for measuring the directness of assistance in LLM tutoring responses - The article operationalizes scaffolding as a five-level ordinal scale characterizing LLM responses by the amount of assistance they provide: L0 Minimal, L1 Prompting, L2 Hinting, L3 Explaining, and L4 Solving, where \"Higher levels provide more directly usable assistance and leave less of the task fo
Five-level scale of Active Student Participation (ASP) in whole-school intervention activities - The review proposes an organizing structure that grades program activities on a 5-point scale of active student participation, partly following the school participation scale of the HBSC questionnaire (De Róiste et al., 2012).
Five-level taxonomy of argument quality for coding reasoning gains - The review codes argumentative reasoning gains on five levels combining structural and functional elements: main (claims/reasons), elaborated (evidence, questions, TAP structure), complex (internal relationships and quality of evidence/justification), dialogic (transactivity, critique, rebuttal), an
Five-part question taxonomy with multi-endpoint metrics for evaluating course-grounded AI tutors - The article extends a prior 30-question generic/specific set into a five-part taxonomy of classroom question types: \"(a) single fact lookup, (b) multi hop concept linking, (c) thematic explanation and synthesis, (d) contradictionandambiguityhandling(Houetal.2024),and (e) curriculum update questions
Five-part scope framework for post-exercise automated team assessment in TTXs - The article defines its assessment scope through five grounded characteristics: assessment targets teams rather than individuals, since a team is the fundamental unit and succeeds only when members collectively resolve tasks; it is summative, using milestones as predefined checkpoints that serve as
Five-pattern taxonomy of AI-use learning behavior - The article proposes a framework characterizing how students interact with an AI assistant during assignment work, distinguishing limited engagement from \"proactive and critical engagement.\" Annotators labeled chat events for prior attempt, request purpose, and answer modification, then aggregated t
Five-step SAS-based item fit analysis procedure for MAP Growth K–2 items - A psychometric quality-control procedure for validating Rasch-calibrated items after difficulty adjustment.
Five-theme, sixteen-indicator framework for coding spatial skills in early childhood curricula - The study coded curricular documents with a deductive six-phase latent thematic analysis in which \"Five themes with a total of sixteen indicators were identified based on previous literature.\" The framework organizes the curricular representation of spatial skills — abilities underpinning processing
Five-type taxonomy of research-integrity learning outcomes - The article organizes RCR training effects into five distinct learning outcome types: knowledge, attitude, sensitivity, judgment, and behavior.
Formula-based estimation of value-added classification error rates using ordinary least squares and empirical Bayes estimators - The article's methodological framework estimates likely error rates for value-added performance classifications using \"formulas based on ordinary least squares and empirical Bayes estimators.\" It applies these formulas to student test score gain data to quantify how often teachers or schools would b
Four AI model families as complementary lenses on educational cognition - The review organizes AI approaches to modeling educational cognition into four families treated as different lenses rather than a formal taxonomy: neural networks, reinforcement learning agents, cognitive architectures, and large language models.
Four levels of functional specificity of theoretical propositions, from non-causal association to functionally specified causal effects - The article proposes that propositions differ in specificity, which determines their falsifiability.
Four measurable properties for evaluating adaptive virtual patients - The article proposes four measurable properties an adaptive virtual patient should satisfy, each with its own metric: P1 Responsiveness (state changes in the direction the data predicts), P2 Gradual Build-up (the state \"climbs steadily across the session and only as the user earns it\"), P3 Faithful
Four observable group-level emotion regulation strategies coding scheme - The article distinguishes four qualitatively different, socially observable emotion regulation strategies used by groups during collaborative tasks.
Four options for accounting for co-teaching in value-added models - The paper organizes the policy problem of co-teaching in value-added models around four named methods: \"the partial credit method, the teacher team method, the full roster method, and the full roster-plus method.\" Each option represents a different way of assigning credit for jointly taught students
Four paradigms of shared knowledge span individual representations, shared objects, shared situations, and shared communities - The editorial organizes CSCL analyses of shared knowledge into four paradigms spanning a range from individual to community levels.
Four shifts for establishing ground truth in AIED: diagnostic IRR, transparent reporting, LLM risk mitigation, and validity-first evidence - The article's central framework proposes four practical shifts for establishing ground truth in AIED labeling: (1) treat IRR as \"a diagnostic signal to localize disagreement and refine constructs rather than a mechanical acceptance threshold\"; (2) require transparent reporting of rater expertise, co
Four-activity framework of online collaboration - The article proposes that online collaboration comprises four types of activities: task-related activities in the content space (exchanging information, asking questions), social activities in the relational space (greetings, social support, grounding via shared understanding), regulation of task-re
Four-capability taxonomy for organizing educational foundation-model supervision - OmniEdu organizes education-specific supervision around four complementary capability categories defined \"according to the behaviors to be learned\" rather than by source or subject.
Four-category coding scheme for peer assessment designs - The review defines peer assessment as \"a learning phenomenon where individuals or social constellations (e.g., pair, group, team, community)—within a physical and/or virtual environment—exchange, react to, interact and/or act upon information about individual performance and/or individual contributi
Four-category framework for evaluating the generalizability of learning techniques - The monograph organizes each technique review around four categories of variables, inspired by Jenkins' (1979) model: \"materials, learning conditions, student characteristics, and criterion tasks.\" Materials pertain to the content to be learned; learning conditions concern how and where the techniqu
Four-category functional classification of student GenAI use (Learning, Formatting, Doing Task, Feedback) - A descriptive coding framework developed inductively from students' mandatory GenAI-use declarations in the physics unit, classifying the \"clearly stated pedagogical function of each reported GenAI use\" into four themes: Learning (researching, understanding, conceptualising, refining understanding),
Four-category functional taxonomy of AI tools shaping collaborative learning in higher education - The review organizes AI tools by underlying functionality into four categories: predictive analytics (machine learning models, early warning systems) that predict student success and identify at-risk students; language-based systems (chatbots, large language models, natural language processors) enab
Four-category taxonomy of 12 preschool SR/EF intervention approaches - The review organises preschool SR/EF interventions into a taxonomy building on Takacs and Kassai (2019), in which \"four higher-order categories comprised of 12 distinct intervention approaches were identified\": Play (mediated structured play, semi-structured creative play), Social-Emotional (mindful
Four-category taxonomy of classroom management strategy foci: teacher, student behavior, social-emotional development, and teacher–student relationships - The meta-analysis organizes classroom management programs by coding each included intervention for the presence or absence of four categories of strategies.
Four-characteristic framework for valid formative assessment - The review presents validity as a matter of degree and identifies characteristics that underpin valid formative assessment.
Four-characteristic model of national historical master narratives - The article presents a model distinguishing four characteristics of national historical master narratives: the historical subject established by inclusion and exclusion, identification processes attaching affect and value judgments, historical events simplified around one common narrative theme, and
Four-component validity scoring scheme for eye-tracking measures - A structured scoring system adapted from Cook et al.
Four-condition dichotic listening paradigm with mTRF speech-envelope reconstruction - A child-appropriate EEG paradigm in which participants attend to an English narrative in one ear while ignoring interference in the other across four conditions: Single talker (no interference), English–English, English–Latin, and English–Musical Rain.
Four-criterion selection of neuropsychological measures for portable classroom EEG - The article proposes screening candidate neuro-psychological measures against four criteria: potential to add unique information about challenge and threat states, compatibility with portable EEG devices, a strong empirical and theoretical basis for inference from EEG data, and the ability to remain
Four-dimension learning analytics reference model (what, how, why, who) - The review adopts Chatti et al.'s (2012) reference model in which \"LA consists of four dimensions including (1) what (what types of data does the system capture and analyze), (2) how (how does the system perform analytics), (3) why (for what reasons does the system gather and analyze data), and (4)
Four-dimension team creative performance framework for DST evaluation - The study adopted Liu et al.'s (2023) TCP assessment framework comprising four dimensions: \"novelty, relevance, user experience, and sustainability.\" Novelty assesses originality and innovation; relevance focuses on practical applicability; user experience evaluates engagement through authenticity,
Four-function taxonomy of pictures in multimedia testing (decorative, representational, organizational, informational) - The review organizes pictures in test items by their functional relationship to the accompanying text, adapting Levin's MML classification.
Four-layer model of pedagogical safety for educational RL with the Reward Hacking Severity Index - The article formalizes pedagogical safety as four constraint classes: \"Structural Safety (C1):Prerequisites are architecturally enforced via action masking\", progress safety (C2, windowed mastery gain), behavioral safety (C3, minimum cognitive demand), and alignment safety (C4, engagement coupled to
Four-level response safety classification for evaluating educational LLM safety - EduZone grades every model response into one of four ordered safety levels rather than a binary safe/unsafe decision.
Four-level rubric for teachers' reflective prompts adapted from Van Manen's reflective levels - The study adapts Van Manen's (1977) cumulative levels of reflection into a four-level rubric for classifying teachers' reflective prompts in learning modules.
Four-metric behavioral taxonomy for distinguishing authentic iterative coding from automated workflows - The article defines four analytical metrics derived from explicit API event logs to \"differentiate authentic learning behaviors from automated or externally synthesized workflows\": (a) a class rhythm heatmap of temporal interaction density, (b) cumulative platform engagement showing staircase-like g
Four-process framework for collaborative problem solving in ill-structured STEM tasks (P1–P4) - The article operationalizes Ge and Land's four problem-solving processes for ill-structured engineering tasks: P1 (exploring and representing the problem, including joint FBD construction), P2 (planning how to solve, considering multiple approaches), P3 (attempting to solve with justifications), and
Four-requirement framework for education-facing LLMs - The article organizes evaluation around four requirements an education-facing model should jointly satisfy, drawn from the learning-sciences literature (Shulman's pedagogical content knowledge, Vygotsky's zone of proximal development, Bloom's tutoring ideal, and AI-in-education ethics frameworks).
Four-role taxonomy of GenAI roles: instructor, enhancement tool, replacement tool, collaborator - The article develops a four-category taxonomy of the functional roles students assign to GenAI in authentic interaction.
Four-stage dataset preparation workflow for assessment research - A reproducible preparation pattern for consolidating heterogeneous examiner files into an answer-level dataset.
Four-stage dropout timing definitions for cross-institutional persistence comparison - The article collapses Simpson's nine potential dropout stages into four timing-based definitions of dropout (Table 2), distinguishing drops before course start, during the initial or Orientation week, and during instruction, and counting only \"active drops\" while excluding dormant (MIA) students.
Four-step methodological framework for measuring LLM alignment to intended impact in high-noise contexts - The article proposes a general framework for evaluating whether LLM judgments align with real-world outcomes when experimental control is infeasible.
Four-type taxonomy of behavioural change in team learning behaviours: dynamics, pattern, development, and minor change - The review defines four types of change in team learning behaviours, distinguished by the shape of the sequence of increases and decreases.
Four-type taxonomy of computer-assisted word reading interventions - The article organizes computer-supported word reading interventions into four program types: foundational (comprehensive) reading instruction, supplementary alphabetics, reading fluency programs, and remedial reading programs for at-risk students in higher grades.
Four-type taxonomy of word-problem process models by process order and mental representation - The article organizes existing cognitive models of word-problem solving along two axes: whether text comprehension and computation occur in a \"linear\" or \"flexible\" order, and whether they use \"separable or integrated mental representations\".
Framework of computer-based scaffolding design characteristics and moderators - The article organizes computer-based scaffolding research along several coded dimensions: context of use (problem-based, case-based, design-based, project-based, inquiry-based, modeling/visualization, problem solving), scaffolding change (none, fading, adding, fading/adding), scaffolding logic (none
Framework of reliability and validity evidence categories for classifying RCR outcome-measure validation - The review develops a coding framework distinguishing forms of reliability evidence (test–retest, parallel forms, split-half, internal consistency via Cronbach's alpha or Kuder–Richardson formulas) and validity evidence (content validity including cognitive interviews and S-CVIs, face validity, crit
Framework of teachers' judgments of students' monitoring skills with three accuracy measures - The article introduces a framework that integrates teachers' judgments of students' performance, teachers' judgments of students' monitoring judgments, and students' actual performance and monitoring judgments.
FSM-based evaluation of interactivity in AI-generated educational materials (EE-Eval) - The article conceptualizes interactivity as \"a finite space of learner-controllable states and transitions\", represented as a directed graph G = (V, E, M) where V denotes semantic interaction states, E transitions, and M task-level metadata such as pedagogical intent.
Functional system incorporating tools as the unit of analysis for CSCL microanalysis - Arnseth and Krange argue, building on Vygotsky and Wertsch, that \"the unit of analysis needs to be a functional system incorporating tools.\" This stance opens up understanding of the multiplicity of activities, in which collaboration between students and computational tools can take multiple paths,
Fundamental-versus-peripheral hypothesis taxonomy for model replication and falsification - To answer criticisms that social-cognitive models are unfalsifiable, the authors classify each of the model's 23 hypotheses as fundamental or peripheral.
Funds of Knowledge - The digest's recommendation set for teachers who plan similar projects: \"the following minimal conditions are suggested\" — theoretical teacher preparation, believing households contain abundant social and intellectual resources; home visits as participant-observers entering as learners; teacher labs
Future Technology Workshop (FTW) method - The Future Technology Workshop (FTW) is a structured participatory method in which participants with everyday knowledge of a domain envision and design interactions between current and future technology and activity.

G¶

Game-Based Assessment (GBA) model - The GBA model is a design strategy developed by the Games, Learning and Society group to capture evidence of player learning during game play rather than treating the game as a black box.
Gap-analysis method for identifying gateway courses via predicted versus observed DFW rates and persistence gaps - A learning-analytics method using institutional student records.
General polytomous testlet model - The general polytomous testlet model is an item response theory model for testlet-based tests containing both dichotomously and polytomously scored items.
Gesture Memory Support Model (GMSM): seven dimensions for characterizing instructional gestures - The GMSM is a preliminary theoretical and operational framework for characterizing instructional gestures in L2 vocabulary learning.
Grade productivity profile - The guide introduces a value-added framework in which each school grade's productivity profile presents \"three inter-related pieces of information for each school grade.\" Input status is the previous year's matched test scores, output is end-of-year scores for the same stable students, and the gain
Graduated assessment rubric separating answer correctness from method validity - The rubric adapts the mathematical assessment distinction between the answer mark (correct output) and the method mark (valid reasoning) to automated diagnosis.
Graesser et al. question taxonomy adapted to student-AI programming interactions - The article adapts the Graesser et al.
Graph-theory-based passage equating for WCPM scores - The report introduces a method that models passages as nodes and same-student passage pairs as edges, then uses graph theory to choose a reference passage with the most edges, adequate sample sizes, and medium difficulty, and to define the \"shortest path\" from each passage to the reference passage s
Graphical model framework for MML estimation of multidimensional IRT models - The article presents a framework in which a multidimensional IRT model is represented as a directed acyclic graph whose nodes are random variables and whose edges encode conditional (in)dependence relations.
Group disciplinary engagement (GDE) framework - The GDE framework conceptualizes engagement at the group rather than individual level, extending productive disciplinary engagement (Engle & Conant) with a situative perspective.
Group-plus-individual (person-oriented) analysis of intervention outcomes - This framework holds that evaluating remedial interventions solely by group-level improvement can obscure subgroup effects, so outcomes should be examined \"at both the group and individual (or ''person-oriented'') level\".
GSR model: realism as three independently manipulable dimensions of geometry, shading, and rendering - The geometry, shading, and rendering (GSR) model is a technical taxonomy of realism grounded in the actual production process of computer-generated imagery in 3D software.
Gunawardena et al.'s Interaction Analysis Model (IAM) for collaborative knowledge construction - The IAM describes CKC as evolving \"from lower mental functions, i.e., 'pooling of knowledge,' to higher mental functions through five distinct CKC phases\": (1) sharing and comparing information, (2) discovering and exploring dissonance or disagreement, (3) negotiating shared meaning or co-constructi

H¶

HARMOGEN-R: a two-stage hybrid framework for AI rubric generation with reasoning-enhanced generation and low-temperature synthesis - HARMOGEN-R (Hybrid Assessment Rubric Model Generation with Reasoning) is a two-stage rubric generation pipeline.
Hattie and Timperley's feedback model - The article adopts this theoretical framework to analyse the effectiveness of textual feedback.
Henmon's P.E. sentence-difficulty scaling method - The article scales each sentence's difficulty by the same method used for vocabulary words: given the per cent of pupils translating it correctly, its position is located on the base line of the normal frequency surface in P.E.
Hierarchical causal models representing both specific theories and framework theories - The article argues that recently developed hierarchical causal models can capture the theory theory's claim that specific causal theories are embedded in more abstract framework theories.
Hierarchical framework of domain-specific attitude constructs for synthesizing attitude outcomes - The article proposes a hierarchical conceptual framework (Fig.
Hierarchical resampling simulation framework for evaluating metacognitive measures - A simulation methodology in which real confidence-judgment datasets are resampled hierarchically (participants with replacement, then trials within participants) to create realistic simulated experiments without committing to a mathematical model of confidence.
History-conditioned student simulation: two-component profile-generator-plus-simulator framework - The article establishes the task of history-conditioned student simulation, where a model predicts a student's next tutoring-dialogue turn conditioned on their prior question-answering and dialogue history.
Homogeneous-outcome selection for causal learning-loss analysis - This design principle holds that causal estimates of instructional disruptions require tight confinement of the counterfactual trend.
Human-validated LLM reconstruction of syllabus assessment weights - The pipeline converts syllabus PDFs to Markdown and uses an LLM through a university-managed API gateway to reconstruct final-grade weights for each assessment category, which feed the continuous GenAI-susceptibility treatment variable.
Hybrid confidence framework integrating model-based confidence with dataset-derived aleatoric uncertainty via semantic heterogeneity - The framework treats LLM grading unreliability as arising from two sources: epistemic uncertainty, estimated from model-based signals (verbalized, latent, and consistency-based confidence), and aleatoric uncertainty, estimated from the dataset.
Hypothesis-testing framework for computing error rates of value-added performance measurement systems - The report frames teacher and school performance measurement as a hypothesis-testing problem, in which a decision (for example, whether an educator's performance differs from the average) can be a true positive, false positive, true negative, or false negative.

I¶

ICAP-CG lesson-task engagement coding protocol - The ICAP-CG is an observation protocol developed in prior work (Vosniadou et al., 2023) that codes lesson tasks minute-by-minute into Passive, Active, Active/Collaborative, Constructive, and Interactive engagement modes based on the verbs teachers use in task instructions.
Idiodynamic method - The idiodynamic method tracks moment-to-moment changes in subjective emotional intensity during communication events.
Immediate-versus-delayed subjective cognitive load measurement - When using subjective rating scales to measure cognitive load, researchers should not assume that the average of ratings given after each task equals a single rating given after the whole learning phase.
Implementation research within impact studies of education interventions - The guide is a practical resource for researchers who conduct implementation analyses as part of impact studies of education interventions.
Improvement Rate Difference meta-analysis - This element is the meta-analysis itself: the authors synthesized \"Twenty-four single-case studies\" on aided AAC with individuals with ASD \"via an effect size measure, the Improvement Rate Difference (IRD).\" The IRD is used to quantify improvement across single-case designs, a literature base the au
Independent replication of educational intervention studies - The editor argues, quoting Wainer, that replicability is the principal epistemological tenet of modern science: small initial studies tolerate false positives, and other investigators should rush to replicate so anomalous results are brushed aside.
Individual longitudinal growth-curve method for studying mental and physical development - Baldwin advocates the individualizing method: following the same children consecutively over years with both physical and mental measurements, rather than cross-sectional snapshots.
Inferentialist epistemology for analyzing students' mathematical reasoning - The article develops an epistemological framework based on Brandom's inferentialism for studying students' reasoning on the order relation of integers.
Information-based rapid-guessing threshold methods - The article introduces \"A pair of information-based methods\" for setting rapid-guessing thresholds.
Ingredients method for educational cost analysis - The report applies the ingredients method (Belfield et al., 2018), a methodology that involves \"systematically identifying all the resources necessary for implementing\" a program, pricing each component, and calculating total and per-participant costs.
Instructional Quality and Implementation Coding system combining Wood's contingent teaching with Meichenbaum and Biemiller's self-direction dimension - A coding system the authors developed to measure explicit instruction and scaffolding of purpose setting at fine grain.
Instrumental genesis - Instrumental genesis, developed in France by Pierre Rabardel and colleagues, describes how the agent-artifact connection evolves over time through usage and re-design.
Integrated analytical framework combining the technology-based learning framework with Pedaste et al.'s five-phase inquiry cycle - The review adapts the technology-based learning framework (originated by Hsu et al., 2012, with six analytical dimensions: learning devices, learning domains, learning strategies, research methods, participants, and research issues) and integrates Pedaste et al.'s (2015) five-phase inquiry framework
Integrated framework linking cognitive ownership, mutual evaluation, and outcome-aligned verification for assessing AI-assisted work - The article develops a conceptual framework connecting cognitive ownership, mutual evaluation, and selected verification performances so that institutions can make bounded claims about learner capability.
Integrative review - The article itself is an integrative review: it states \"This paper presents an integrative review of theory and research on the broad range of educational interventions to counter community violence.\" Motivated by \"alarming rates of community violence and the detrimental impact on youth and communit
Interaction geography - Interaction geography is an analytical approach introduced by Shapiro, Hall, and Owens that connects \"novel forms of analyses\" to constitute \"the key elements of the analytical approach.\" It extends interaction analysis by including physical movements: through movement analysis researchers \"can expl
Interactional analysis of emergent distributed leadership in collaborative learning - The editorial argues against folk-theory conceptions that label individuals as leader, follower, lurker or lazy, and against simply assigning a leadership role to one student per group.
Interactional Ethnography and dialogic intervisualizing - The editorial describes Bridges et al.'s methodological lens, Interactional Ethnography (IE), as \"a logic-of-inquiry that can be used to examine collaborative processes\" as undergraduate students and their facilitator work with multimodal text.
Interpersonal multilevel analysis of regulation in collaborative learning - The editorial reports that Su et al.
Intersectional disaggregation of school-climate survey data - This report's organizing method is to disaggregate one large survey dataset along intersecting identity axes rather than analyzing LGBTQ students as a single group.
IRT-based repeated-measurement linking with re-randomized groups and concurrent links - The article recommends an IRT framework for its repeated measurement linking design because IRT \"easily allows for incomplete designs and for equality constraints between item parameters.\" The author notes that when new randomized groups are formed at each occasion and all links are put to work in c
Item characteristic curve (unsigned area) method for DIF detection under item response theory - The ICC method compares item characteristic curves estimated separately for reference and focal groups, computing the unsigned area between the male and female curves as a DIF index, using a probability index developed by Linn.
Item response time analysis as response-process metadata - The article treats item response time as one common type of test metadata generated by computer-based assessment.

J¶

Jackknife leave-one-year-out empirical Bayes estimation of teacher grade effects - A two-step value-added estimation procedure, adapted from the teacher value-added literature, that produces teacher grade effects used to predict student outcomes.
Jan van Aalst's seven-code discourse coding scheme - A coding scheme developed to analyze asynchronous group discourse in Knowledge Forum and to identify mechanisms explaining between-group differences in summary-note performance.
Joint GLMM model for IRT estimation with collateral information - The article applies a multivariate joint model under the generalized linear mixed models framework that \"simultaneously models binary item response and continuous RT response on test items that nested within person\" (Klein Entink, Fox, & van der Linden, 2009).
Joint modeling of English proficiency growth and reclassification timing (SREM) - Because the joint model outperformed the conventional approach, the article supports designing EL policy analytics that estimate proficiency growth and reclassification timing together rather than separately.
Jonassen-informed prompt-variation framework with MAPS-based evaluation for probing LRM physics reasoning - The article adapts Jonassen's problem-classification framework to generative-AI problem solving along two dimensions: \"prompt-specificity, or how much information is explicitly given versus left for the model to infer, and modality, or whether the problem is presented through text alone or with visu
Justified-claims coding scheme for self-explanation quality in ill-defined statistics problems - The article develops a coding scheme assessing self-explanation quality as the degree to which written problem solutions include correctly applied and justified conceptual information.

K¶

KSA three-axis evaluation framework with a four-level Skills hierarchy (Center→Role→Scenario→Sub-scenario) - The framework operationalizes educational LLM capability using the KSA (Knowledge-Skills-Attitude) model from educational assessment.
KT-grounded simulated learner - The simulated learner couples two components: a knowledge tracing model (AKT) that maintains a per-KC belief map trained on real student interaction data, and an LLM that role-plays the student and generates free-response attempts.

L¶

L* statistic - L* is a modified version of D'Mello's L statistic intended for use when self-transitions are excluded.
Large-scale randomized controlled trial design for evaluating supplemental literacy curricula - The study's organizing method is \"a large-scale randomized controlled trial of the effects of four supplemental reading comprehension curricula.\" Impact analyses are conducted separately by study year, allowing effects to differ across years and across teacher subgroups defined by experience using t
Latent change score (LCS) modeling framework for comparing developmental theories - The article uses bivariate latent change score (LCS) models, which \"conceptualize differences between successive measurements as latent change factors,\" allowing within-subject change to be modeled via self-feedback (β) and coupling (γ) parameters.
Latent structural readiness space for digital and AI transformation - The framework trains a variational autoencoder on standardized country-level indicators of autonomy, digital skills and teacher support, projecting them into a two-dimensional latent space interpreted as \"a structural readiness space for digital and AI-related transformations in education\".
Latent-score linking method - A prediction-model approach that links scores across different K-2 literacy assessments using test records already available in administrative data.
Learning Study model - The Learning Study is \"a fusion between lesson study and design experiment\" in which teachers and a researcher jointly plan a lesson around a chosen learning object, implement it with one pupil group, and analyze pre-test, post-test and delayed post-test results together with video-recorded lessons.
Learning-issue–objective overlap measure - One of two procedures developed at the University of Limburg to reflect \"the extent to which students are able to identify important learning issues given a particular problem\".
Linearity-assumption testing for mental-effort relations - The article recommends that researchers check whether relations involving self-rated mental effort are linear before fitting and interpreting linear models, because \"predictions based on linear models may be substantially biased for most of the range of mental effort\".
LLM scoring, generated-feedback, and essay-feature analysis framework - The article proposes an analysis framework in which \"Essays are first evaluated by LLMs to produce both predicted scores and textual feedback.\" Essay-level features (length, Flesch Reading Ease readability, and LanguageTool error counts) and feedback-derived signals (trait-specific praise and critic
Lonchamp's three-level integration of dialog, knowledge, and action analysis for dual-interaction CSCL environments - Jacques Lonchamp proposes a systematic integration of three analytic approaches, describing \"three levels of analysis — dialog, knowledge, and action — which he claims fit together like semantics, syntax, and pragmatics to provide an integrated view of communication.\" He applies this within dual-int
Longitudinal achievement growth analysis - This principle holds that educators and researchers should use longitudinal data analysis to track how academic achievement for DHH students progresses over time.
Looking-while-listening procedure - The looking-while-listening procedure assesses efficiency of online spoken language comprehension in infants and toddlers.
Loss-function decision rule for late pretest use in education RCTs - Education RCTs often use pretest-posttest designs because including pretests improves the precision of estimated treatment effects.
LT-MKT framework: joint modeling of cognitive load and knowledge transfer for multi-domain knowledge tracing - LT-MKT is a knowledge tracing framework for multi-domain learning scenarios that jointly models two factors the authors argue are overlooked by single-domain methods: cognitive load and knowledge transfer.

M¶

Machine-learning prediction of accessible assessment-item quality - The researchers introduced three word embedding methods and three classifiers as a toolkit for predicting item quality for accessible assessments.
Making-Process-Rug video-analysis method - The Making-Process-Rug is a visual video-analysis method developed by the authors to trace \"social-discursive, materially mediated, individual, and collaborative making processes\" across a longitudinal co-invention project.
Many-Facet Rasch Model - MFRM, developed by Linacre in 1989 as an extension of the partial credit model, is an IRT model in which \"all so urces of variability , such as raters, items and persons, which have the potential to influence measurement results , are considered simultaneously\".
MAP Growth item calibration workflow - The existing tool, built in T-SQL within the Growth Research Database, runs an automated pipeline: data cleaning, item parameter estimation via a brute-force grid search over RIT values 100-350 that minimizes mean square fit with person abilities fixed, and a Model of Man (MoM) procedure that predic
Marginal structural models with inverse probability of treatment weighting - The article adapts a counterfactual framework for time-varying treatments in which \"Inverse probability of treatment (IPT) weighting is an alternative approach specifically developed to adjust for confounding by time-varying covariates.\" A marginal structural model specifies the logit of graduation
Matrix sampling with planned missingness in large-scale assessments - Large-scale assessments organize items into testlets counterbalanced across students within a sampling frame, so that 'no student ever takes the entire battery of assessments, or even all the items on any one test'.
MBP-KT - MBP-KT is a three-component framework for knowledge tracing.
MeDOL framework for reviewing learning-to-argue instructional approaches - The MeDOL framework is the review's organizing scheme, adapting Morrison et al.'s instructional effectiveness model to the diversity of LTA research.
Meier et al. rating schema for productive collaboration - The study's model of productive collaboration is the rating schema by Meier et al.
Meijer et al. taxonomy of 56 metacognitive text-studying activities - A taxonomy for coding think-aloud protocols of secondary school students studying a history text, developed by Meijer et al.
Mental-age score conversion and norming method - The article details procedures for reducing heterogeneous survey data to the common basis of mental age.
Messick's six-source unified validity framework for AI evaluation - The article anchors its validity discussion in Messick's unified theory of validity, which holds that validity is a property of inferences and decisions based on scores, not of instruments.
Meta-d′ as a performance-normalized measure of metacognitive efficiency - Meta-d' is a signal detection theoretic measure of relative metacognitive accuracy, attributed in the article to Maniscalco and Lau (2012), that explicitly models the relation between primary task performance and metacognition.
Methodological parameter framework for designing intensive longitudinal studies - The review organises the design of intensive longitudinal studies around core methodological parameters: study duration, assessment frequency, sampling scheme (fixed, random, semi-random, event-based), item phrasing and questionnaire density, and data-collection device.
Methods Reproducibility framework for meta-analysis: four levels from Computational Reproducibility to Outcome Reproducibility - The article introduces a framework targeted at facilitating the reproducibility of meta-analysis, organized in four levels building on the TOP-guidelines, the Reproducibility Enhancement Principles, and Lakens et al.'s recommendations.
Metzler's three lines of experimental investigation for geometry pedagogy - The article organizes experimental work on geometry teaching around three questions: whether the desired results are obtained under a given method, what the best method of obtaining them is, and whether obtaining them transfers to other subjects.
Micro-ecological approach to cross-level CSCL analysis - Borge and Mercier propose this approach as an alternative to the four overarching CSCL learning theories (information-processing, constructivism, socio-constructivist/socio-cultural, group cognition), arguing these lenses are \"too narrow to capture the true contextual richness of CSCL in action.\" It
Microscopic network measures as indicators of sentence-level integration during reading - The article frames reading networks with sentences as nodes and transitions between sentences as weighted edges, and proposes four microscopic measures of integration.
MISCA model (Message, Implementation, Student Characteristics, Context, Agents) - The review adopts the MISCA (Message, Implementation, Student Characteristics, Context, Agents) model proposed by Panadero and Lipnevich (2022), which integrates five common components for classifying feedback practices: Content, Function, Student Characteristics, Presentation, and Source.
Mixed functional data analysis for grade-varying longitudinal predictor effects - The article applies longitudinal mixed functional data analysis (FDA) to estimate regression effects as functions of grade, using linear splines for fixed and random effect functions.
Mobile eye-tracking methodology for capturing joint visual attention in co-located groups - The paper describes a methodology for studying co-located collaborative groups using mobile eye-trackers, providing \"a comprehensive description of our data collection and analysis processes so that other researchers can take advantage of this cutting-edge technology.\" Data were collected from 27 st
Model of collaborative knowledge building across interconnected planes - Figure 3 presents \"A model of collaborative knowledge building\", adapted from Stahl 2006 Ch.
Model-comparison approach for linear vs non-linear mental-effort relations (Models 0–4) - A reusable analytic procedure the article introduces: five competing two-level mixed-effects models (random participant intercepts) ordered by complexity — no relation, linear, quadratic, cubic, and more complex than cubic (mental effort as factor) — compared on marginal R2, AIC and BIC.
Multi-angle video capture for human–material analysis - This capture pattern combines eye-level, bird's-eye-view, and close-up turntable videos so that material movements are analytically visible.
Multi-armed randomized trial analysis - The report's guidance is that analysts running studies with several intervention arms should not simply reuse two-group methods: estimators \"need to be modified for the multi-armed design,\" hypothesis tests across pairwise contrasts require multiple comparison adjustments, and CACE estimation demand
Multi-dimensional coding procedure for small group chat communication - A multi-dimensional coding procedure for analyzing small group chat communication in CSCL research, developed by the authors to address three methodological issues: \"the unit of analysis and unit fragmentation\", \"the reconstruction of the response structure\" and \"determining reliability without over
Multi-dimensional fidelity framework for LLM student simulation - The article names competency bias as \"the inherent tendency of preference-tuned models to produce correct solutions even when prompted to simulate novice students\", and frames simulating learners as a multi-dimensional fidelity problem with behavioral, epistemic, and perceptual components.
Multi-method research framework combining SNA, content analysis and contextual analysis for networked learning - The article develops a framework in which three methods answer complementary questions about networked learning: social network analysis to find out \"who is talking to whom\", content analysis of coded learning and tutoring activities to find out \"what they are talking about\", and contextual analysis
Multi-rater latent growth curve model - The article introduces a model for estimating growth from longitudinal data in which multiple raters evaluate a construct such as a student's self-efficacy.
Multi-turn synthetic-dialogue evaluation protocol for AI tutors - A novel evaluation protocol in which a fine-tuned student model and a candidate tutor model generate synthetic multi-turn dialogues that are scored by conversational statistics and a reasoning LLM judge, with results aggregated.
Multidimensional evaluation framework separating reference-answer similarity, structured-output stability, and domain-specific correctness - The article proposes an evaluation framework for educational question-answering models that combines automatic text metrics with structured-output features and bootstrap confidence intervals.
Multidimensional Recurrence Quantification Analysis (MdRQA) with surrogate-data comparison - Multidimensional Recurrence Quantification Analysis (MdRQA) is the analytic method the article uses to quantify physiological synchrony in multivariate EDA time series.
Multidisciplinary taxonomy of educational approaches to counter community violence - The review establishes a multidisciplinary taxonomy that organizes the broad range of educational interventions to counter community violence into distinguishable approaches.
Multilayered temporal analysis of collaborative-learning roles - This analytical pattern combines several learning-analytics methods into layers to capture role dynamics over time: latent class analysis converts CoI-coded discourse into role states, epistemic network analysis maps code co-occurrence within roles, sequence mining orders roles across weeks, and Mar
Multilevel Analysis of Individual Heterogeneity and Discriminatory Accuracy (MAIHDA) - MAIHDA (Multilevel Analysis of Individual Heterogeneity and Discriminatory Accuracy) is a statistical framework in which individuals are nested within intersectional strata defined by combinations of social categories.
Multilevel conceptualization of autonomy support: climate, contextual, and perceived-difference constructs - The article conceptualizes teachers' autonomy support as a multilevel construct with distinct operationalizations: class-directed support measured with class-referent items forms an L2 climate construct reflecting the \"overall classroom-specific autonomy-supportive climate\"; individual support is an
Multilevel conceptualization of moderators in GenAI-mediated writing research (technological, learner, and contextual factors) - The article proposes a multilevel framework for understanding variability in GenAI-mediated writing outcomes.
Multilevel latent covariate model for unbiased estimation of class-level climate–outcome relations - The article recaps Lüdtke et al.'s (2008) multilevel latent covariate model, in which students' ratings of classroom climate are latently aggregated to a latent class mean that is regressed on an outcome at the class level.
Multilevel latent variable model (ML-LVM) of judgment accuracy - The article introduces the ML-LVM, in which each rater's judgments are modeled with a rater-specific bivariate normal distribution whose mean, variance, and covariance parameters correspond to the level, differentiation, and rank components of judgment accuracy.
Multilevel modeling of teacher self-efficacy and instructional quality across teacher and school levels - The article argues that research on teacher motivation and instruction should not neglect the school context: teachers' self-efficacy should be modeled both as an individual variable and as a school-level composite, and school characteristics (collective efficacy, collaboration, participation, leade
Multilevel modelling of hierarchically clustered CSCL content-analysis data - The article presents multilevel modelling as a statistical approach for hierarchically clustered data in collaborative learning research, in which messages are nested within students (and measurement occasions) within groups.
Multilingual embedding-based automated classification of self-regulated-learning think-aloud utterances - The pipeline vectorizes English and German think-aloud utterances with OpenAI's text-embedding-3-small, \"a pre- trained sentence embedding model that converts textual input into a high-dimensional vector with a length of 1,536\".
Multimodal analysis - Multimodal analysis is one of three methodological advances the report highlights, defined as \"Integrating multiple streams of data, such as audio, video, eye gaze, sensors, and clickstream data.\" The introduction elaborates it as \"the study of learning with data streams that include physical moveme
Multimodal engagement-prediction framework with Perceiver IO fusion, hierarchical Bayesian participant context, and uncertainty-aware heads - A computational framework that jointly predicts a continuous engagement score in [1,5] and a binary label from tutoring-video clips.
Multimodal process-tracing approach to rubric use - The article adopts a methodological framework combining eye-tracking with think-aloud protocols to capture students' real-time engagement with rubrics during academic tasks.
Multiple-design-element interrupted time series evaluation - Because the basic single-group ITS depends on extrapolating the pretest functional form as a counterfactual, an assumption the authors call problematic, evaluators should add multiple design elements to the basic ITS structure.
Multivariate Bayesian Causal Forests - This study adopts a multivariate extension of Bayesian Causal Forests (BCF), a non-parametric causal inference machine learning model built on Bayesian Additive Regression Trees.
Multivariate true score model for achievement and growth norming - The report uses a model-based approach in which observed scores are decomposed into true scores and error scores, with the standard error of measurement introduced so that imprecision of observed scores is accounted for.
Mutualistic network approach to motivation-achievement relations - The review advocates a network approach in which motivation and achievement constructs are nodes and bidirectional partial correlations are edges, so constructs emerge empirically as clusters rather than being assumed a priori.

N¶

N-gram uptake analysis for human–AI co-writing - The article proposes analyzing textual uptake in AI-supported writing via recurring n-grams between ChatGPT output and the writer's final text, extending Bereiter and Scardamalia's knowledge telling/knowledge transforming distinction to human–AI interaction.
Negative priming paradigm adapted to misleading strategies - The negative priming paradigm, originally designed by Tipper (1985), rests on the logic that \"if a distractor (or a strategy) is inhibited on a given item (i.e., the prime), then its activation on the following item (i.e., the probe) should be disrupted\".
Neural-symbolic knowledge tracing with conservative mastery and non-mastery rules - The article recommends encoding pedagogically motivated constraints as learnable-weight symbolic rules that modulate neural predictions when simple evidence patterns occur in the interaction history.
Newell-inspired time-scale taxonomy for aligning SRL measurement to theorized process duration - Adapting Newell's bands taxonomy from cognitive science, the article proposes that SRL phenomena be matched to the time scale at which they are theorized to occur: cognitive-band events over seconds observed with trace methodologies and think-alouds, rational-band events over minutes to hours includ
Nine-step game design methodology for generating a psychological profile of players - The article proposes a game design process dedicated to psychological profiling, stating: \"Following our framework, we propose a game design process in 9 steps\".

O¶

Office-hours queue simulation framework - A simulation framework for comparing office-hours queue scheduling strategies without live experimentation.
Offline evaluation with calibrated LLM judges for tutoring behaviors - The team maintains targeted offline datasets sampled from production completions, including an answer-giveaway dataset where \"The final dataset consists of 63 data points\" and a Tutor Me input-classification dataset where \"the dataset consisted of 794 data points.\" LLM judges calibrated against huma
On-line versus off-line methods for assessing strategy use - The article distinguishes two families of strategy-use assessment.
One-shot GPT-4o-mini codebook-based annotation of LLM math-tutor dialogue - The article reproduces an 82-code educational codebook organised into 7 major categories (MPS, MKT, Cognition, Metacognition, PIU, Student Intent Utterance, Interaction/Discourse) by inlining it into a single GPT-4o-mini prompt at temperature 0: \"we reproduce an 82-code educational codebook on119,00
Operationalization and description of instructional treatments in effect-of-instruction research - The article argues that effect-of-instruction studies fail when instruction is merely given a methods label.
Ordered network analysis (ONA) - Ordered network analysis extends epistemic network analysis by incorporating temporal ordering into network construction, \"allowing edges to represent ordered transitions between codes rather than unordered co-occurrence.\" The article applies ONA to retrospective cued-recall affect reports from a sh
Outcome-oriented equity framework distinguishing access equity from impact equity - The paper distinguishes access equity, whether students and institutions can obtain devices, broadband, software, and support, from impact equity, whether AI-enabled learning produces comparable gains across subgroups.
Overview-of-reviews framework for professional educational and quality assurance behavior change interventions - The article organizes the evidence on changing professional behavior by systematically identifying, appraising, and synthesizing systematic reviews rather than primary studies.

P¶

Partially specified structural causal model (PSCM) protocol for educational assessment - The article proposes modelling educational tests as partially specified structural causal models in which observable endogenous variables (answers and hint usage) are generated by latent exogenous variables.
Pattern-based knowledge component extraction from program ASTs - The framework defines knowledge components as \"recurring, semantically meaningful structural patterns based on important subtrees from program Abstract Syntax Trees (ASTs)\".
PEA (Policy, Enforcement, and Authority) analytic lens - PEA organizes assistance governance around three design questions: what forms of assistance are allowed or restricted (Policy), how those boundaries are implemented (Enforcement), and who can control them during use (Authority).
Pentad-based coding scheme (P, S, T, CS, CL) for classifying authentic CSCL designs - The article operationalizes sociocultural design variation with a coding scheme crossing the pentad facets: Participants (P0 teacher, P1 non-authentic practitioners, P2 authentic practitioners), Setting (S0 classroom, S1 non-authentic settings, S2 authentic settings), Time (T0 educational timeframe,
Per-draw robust linking - The linking step maps each independently calibrated period to a common reference metric by minimizing a response-count-weighted pseudo-Huber (robust Haebara) discrepancy between expected-score curves on anchor items.
Person-centered latent-profile analysis for testing relative universalism of psychological theories - The article proposes testing a theory's relative universalism—functional forms invariant while parameter magnitudes vary—by combining latent profile analysis with within-profile structural equation modeling.
Perturbation-based framework: structured interventions as diagnostic probes of cognitive architecture - The article advances a framework in which interventions that alter physiological or motivational states in predictable ways serve as diagnostic tools for probing the organization of cognitive systems, analogous to lesion studies and pharmacological or stimulation manipulations.
Phantom peer spillover effects - The article adopts the concept of phantom effects, coined by Harker and Tymms, defined as \"misleadingly positive effects of aggregated variables in compositional models — peer spillover effects that are simply an artifact of the inadequacy in the statistical procedures used.\" Two facets of student-l
PISA collaborative problem-solving framework - The editorial describes the PISA framework (OECD 2017) as emphasizing \"the relevance of establishing and maintaining shared understanding for collaborative problem solving\" and being used as the analytic stance in the Bause et al.
Planned Comparable-Group Courseware Impact Study - The article distills lessons from the NGCC evaluation about conducting impact studies: measuring impact requires an objective outcome measure, a comparison group, evidence of comparability, adequate numbers in both groups, and holding other influences constant.
Population-first versus individual-first measurement framework - The article articulates an epistemic framework contrasting two measurement traditions.
Population-level inference with plausible values conditional on assessment design - The article's central recommendation: with large-scale assessment data, 'inferences are only possible at the population-level—and then only when framed as conditional on the assessment design and assumptions—not at the level of individual students'.
Post-hoc fairness evaluation testbed for vendor-controlled early-warning systems - The article built a comparative evaluation testbed simulating an institution's position under vendor procurement: a research replica EWS and six post-hoc fairness methods applied without accessing, modifying, or retraining the underlying model.
Poverty-adjusted value-added school effectiveness rating model - A model that rates each school's effectiveness by regressing average school test scores on school poverty level and standardizing the residuals.
Power planning for mediator–outcome linkage analyses in clustered education RCTs - Because the article develops power formulas specifically for exploratory analyses linking teacher practice outcomes to student achievement outcomes in clustered school-based RCTs, it implies that researchers should evaluate power for these analyses when designing studies.
PPMC assuming posterior normality (PPMC-N) - The article introduces PPMC-N, a posterior predictive model checking method for assessing local dependence in unidimensional item response theory models.
Precision-based sample-size planning for clustered education trials - The article recommends that evaluators discuss \"appropriate precision standards\" and then determine, for each random-assignment design, \"the required number of schools to achieve those standards\" using empirical parameter values.
Prediction versus explanation as two cultures of statistical modelling in educational psychology - The article articulates a framework distinguishing explanatory from predictive statistical modelling.
Prefix-based data augmentation for early prediction in educational sequences - A data augmentation technique in which, instead of sliding windows or cropping, \"we generate all prefixes of each sequence (from length 1 to n), allowing the model to learn to make predictions at any point in a student's progression\".
Principal Trait Analysis (PTA) - PTA is a data-driven methodology that \"produces a representative set of textual descriptions we callprincipal traitsfrom a corpus of human collaborator-AI collaborative conversation sessions.\" It runs four stages: LLM-prompted behavior-observation extraction (generic or theory-lens), embedding-based
Principle-to-indicator mapping as an accountability device for AI-enabled pedagogical systems - The framework translates pedagogical design principles into observable indicators using a compact mapping (Table I) from principle to indicator to observation method: transparency of identity and limits (misinterpretation rate via survey/observation), scaffolding toward humans (transition-to-human p
PRISMA-LLM: a three-layer reporting framework for LLM-assisted systematic reviews - PRISMA-LLM is an evidence-informed reporting framework for LLM-assisted and AI-enabled systematic-review workflows, organized to complement PRISMA 2020.
Privacy-preserving heterogeneous multi-LLM federated inference framework for cognitive diagnosis - A four-stage framework in which heterogeneous commercial LLMs (LLaMA-3.3-70B, GPT-4o-mini, Claude-3-Haiku) independently predict student knowledge states via engineered prompts, add calibrated Laplace noise locally to their prediction outputs, and aggregate with residual correction.
Privacy-preserving linkage and analysis of educational data using privacy-enhancing technologies - The article presents a worked scenario in which a DLP used across several school districts for high school math instruction is linked with state college enrollment and persistence records using secure multiparty computation, producing aggregate correlations between engagement patterns and college ou
Probabilistic peer grading models (PG1, PG2, PG3) - The article introduces a family of Bayesian models in which each submission has a latent true score, each grader a bias (tendency to \"inﬂate or deﬂate her assessment\") and a reliability (precision of assessments after bias correction).
Projection approach combining absenteeism-literature estimates with summer learning pattern analyses - The study's method is a projection framework rather than a direct measurement of pandemic learning loss.
Promotion power measure of school effectiveness - Promotion power is a measure of school effectiveness that the study defines as distinguishing \"a school's contributions to student outcomes from the contributions of the background characteristics of the students it serves.\" It accounts for prior achievement and other background characteristics, and
Propensity score stratification with multilevel models - When estimating the achievement effects of school transfer or school choice from non-experimental data, the article recommends combining propensity score stratification with multilevel models.
ProPRL: property-aware framework for prerequisite relation learning - ProPRL is a framework that learns prerequisite relations over ordered concept pairs in educational knowledge graphs.
Prospective value-conflict analysis for educational evaluation - The article recommends that evaluators apply the value-conflict taxonomy prospectively rather than only retrospectively.
Pseudo-temporal ordering: inclusion relations among snapshot skill sets as a proxy for learning progression - The article proposes a framework in which, absent temporal logs, set-inclusion relations among learners' mastered-skill sets induce a pseudo-temporal axis: a learner whose mastered-skill set is a superset of another's can be viewed as a plausible later state.
Psychometric pipeline for mapping LLM performance onto interpretable proficiency levels and subskill mastery - The article's organizing framework integrates educational measurement methods into LLM evaluation: \"we estimate item difficulty using a two-parameter logistic (2PL) item response theory (IRT) model fitted to responses from 2,050 Chinese EFL learners\", then conduct a Bookmark standard-setting study w

Q¶

Quadratic aptitude-treatment interaction regression model - The article introduces a multiple regression model in which the aptitude variable enters as a continuous predictor and visual design is coded with two dummy variables, dumOpt and dumSub, each carrying a quadratic interaction term with ability.
Qualitative meta-synthesis - The article's organizing approach is a qualitative meta-synthesis: rather than computing a new pooled estimate, it compares and contrasts the assumptions and findings of existing meta-analyses of PBL.

R¶

Race-focused and race-reimaged metatheoretical approaches for studying motivation constructs - The article distinguishes two metatheoretical approaches for educational research: race-focused approaches, which \"center on issues of race and are grounded in racial or cultural theories and constructs\" such as racial identity and stereotype threat, and race-reimaged approaches, which reconceptuali
RACES (Reward-Aligned Consistent Essay Scoring) framework - RACES (Reward-Aligned Consistent Essay Scoring) is a unified AES framework that \"integrates LoRA-based parameter-efficient fine-tuning, reward modeling, and PPO with KL regularization\".
Random-effects meta-analysis - This element is the quantitative synthesis itself: studies reporting effects of physical exercise on motor skills and executive functions in children with ADHD were identified \"through Cochrane, PsycInfo, PubMed, Web of Science databases\" and \"Ten publications were selected.\" A \"Random-effects model
Rapid cycle product efficacy assessment and feedback loop for ed-tech pilots - A district-centric, practitioner-focused process in which university researchers help districts pilot ed-tech products, gather mixed-methods efficacy data across five dimensions (student learning, student engagement, teacher support, teacher satisfaction, administrator satisfaction), and feed synthe
Rapid guessing as a behavioral proxy for test effort - Rapid guessing is a behavioral proxy for test effort used to measure whether students are engaged with a test.
Rapid-guessing detection of test taker disengagement - The page's featured item concerns \"rapid guessing\" used \"to detect test taker disengagement\", with an emphasis on validation and applications.
Rapid-guessing detection using item response-time thresholds - This framework identifies disengaged item responses on computer-based tests using response times—the seconds between item presentation and answer.
Rasch rating-scale analysis for deriving educational survey measures - The authors derive survey measures or scales through Rasch rating-scale analysis, an item response latent-trait model in which \"Survey items are used to define a measure based on the relative probability of a respondent choosing each category on each item.\" Individuals are placed on the scale from t
REC-CBM framework: rubric-aware error-correction concept bottleneck model for trustworthy open-ended grading - REC-CBM is a concept bottleneck model for open-ended grading that integrates three components: a rubric-aware concept encoder that learns concept-specific representations over responses using a concept query bank, an ordinal pairwise calibration objective that preserves ranking structure among rubri
Referent-aligned multilevel analysis of relationship measures - Because relationship items differ in their referents — dyadic (\"I like my teacher\"), plural (\"this teacher likes us\"), or averaged (\"I like my teachers\") — the referent should determine the level at which analysis takes place.
Regression discontinuity design power and validity analysis - The evidence base for this working paper is a set of four previously published education studies whose data were reanalyzed to estimate the statistical power of regression discontinuity designs relative to randomized controlled trials.
Regression discontinuity embedded in a piecewise multilevel growth model for entry-age effects - The paper introduces a modeling approach that combines a regression discontinuity design with a piecewise multilevel growth model, allowing causal estimates of entry-age effects on both initial achievement and subsequent growth rates.
Reinforcement Learning Measurement Model (RLMM) - The RLMM is a psychometric measurement framework for sequential process data that treats a person's inverse-temperature parameter βj as an ability to choose higher-valued actions under a Boltzmann policy.
Reliance calibration as an output-contingent classification of reliance decisions - The review defines reliance calibration as \"an item-level, output-contingent classification of whether a reliance decision was appropriate given AI-output quality\", distinguishing it from trust (a psychological attitude) and reliance (an observable decision).
Reproducible educational LLM guardrails benchmark and evaluation protocol - A controlled holdout benchmark of 480 synthetic queries (369 injection, 111 benign) covering six attack families, with stratified bootstrap confidence intervals, paired McNemar tests, and multi-seed sensitivity sweeps, plus unified head-to-head adapter measurements.
Response Time Effort (RTE) - Response Time Effort (RTE) is a measure of test-taking effort computed from item response times.
Response-time-based detection of test taker disengagement - The chapter describes a capability for identifying test taker disengagement using response time, which underlies its analysis of what happens when test takers disengage and how disengagement should be managed during scoring.
Responsible-DKT - The Responsible-DKT approach is organized as a reusable four-stage pipeline: raw logs are pre-processed into chronological per-student sequences, symbolic training samples are generated as sequence-to-sequence ground facts and queries, a neural-symbolic model integrates embeddings, recurrence, and s
Responsiveness construct for analyzing dialogic discussion posts - The authors introduce responsiveness as a construct for analyzing discussion contributions, defining it as \"the extent and ways one's response takes into account and builds on previous posts, both logically and emotionally\".
Retrospective emulation of evidence-centered design features for learning analytics - Because the full ECD framework cannot be applied retrospectively to existing game logs, the article recommends emulating its useful features: a student model operationalizing the hypothesized cognitive dynamics (here, the 2SM stances) and an evidence model mapping those dynamics onto observable log
Revealed comparative advantage measure of teacher equity impact - The article introduces \"a novel measure of teacher quality—called revealed comparative advantage—that estimates teachers' impact on racial achievement differences.\" A teacher's comparative advantage for Black students is their differential impact on Black students relative to non-Black students; a p
Rothstein falsification test - A falsification test for teacher value-added models, introduced by Jesse Rothstein (2010) and re-assessed in this paper.
Rubric-based pedagogical quality scoring for educational AI explanations - The article defines a composite pedagogical quality score on [0,1] as the mean of six binary criteria applied per response: step-by-step exposition, correct mathematical notation, worked example, query coverage, explanation depth, and proof-step granularity.
Rule Space model - The Rule Space model (Tatsuoka, 1983) is an IRT-based diagnostic classification framework in which \"Items are decomposed into attributes, which represent the latent traits that the items assess.\" From an examinee's response vector the model \"infers the most likely combination of attributes the exami

S¶

SAGE taxonomy of four recipience processes targeted by feedback interventions - The SAGE taxonomy is a data-driven framework of four distinct recipience processes that feedback interventions target: Self-appraisal (making judgments about oneself, one's traits, or one's behavior), Assessment literacy (understanding the grading process and applying it to make academic judgments),
SAVVY content-derived visual and auditory information density metrics - Grounded in the Limited Capacity Model of Motivated Mediated Message Processing (LC4MP), SAVVY quantifies information density of both visual and auditory channels per slide-transition-induced event segment.
SC2R: semantics-constrained counterfactual recourse framework for educational decision support - SC2R is a three-component pipeline for educational decision support: a predictive component estimating next-assessment success, an integer-programming recourse component generating candidate intervention plans over a discrete action space, and a semantic validation layer using SHACL.
Scaffolding types arranged on a continuum of self-regulation of problem-solving - The article proposes ordering scaffolding techniques by the degrees of freedom they leave the learner: \"we suggest locating the scaffolding types on different positions on a scale of self-regulation of problem-solving, with examples located from a rather low level followed by role assignments and pr
Scanpath theory of gaze as read-out of a cognitive model - Scanpath theory holds that a scanpath is \"a 'repetitive sequence of saccades and ﬁxations, idiosyncratic to a particular subject [person] and to a particular target pattern'\" and that scanpaths reflect \"the read-out of the internal representation of pictures, the so-called 'cognitive model'\".
SciMAT-based integrated bibliometric and thematic analysis - A research artifact and method: the study \"combines bibliometric performance and thematic analysis of the principal instructional leadership (PIL) research field using SciMAT.\" The analysis \"included 695 articles indexed on two reputable databases, Scopus and Web of Science (WoSCC)\" and covered \"thr
Seasonal learning analysis - The brief organizes its inquiry around seasonal learning patterns: comparing how racial and ethnic achievement gaps change across school-year and summer periods to determine when inequality grows.
SEL status-and-growth early-warning modeling - This framework treats social-emotional learning constructs (academic growth mindset, self-efficacy, self-management) as developmental processes with their own growth trajectories, asking \"whether a student's mean SEL score in 6th grade (status) or growth between 6th through 8th grade is more predict
Self-management intervention taxonomy for academic outcomes - The review organizes academic self-management interventions for students with emotional and behavioral disorders into five technique types, reporting that \"Self-monitoring interventions were the predominant type of self-management technique used by researchers.\" The taxonomy distinguishes self-evalu
Self-referential analytic assessment framework - An evaluation paradigm for analytic essay scoring that shifts from normative (inter-learner ranking) to within-learner interpretation of analytic profiles.
Self-reflective LLM annotation agent with iterative criteria refinement and short-term memory - An LLM-based agent that labels each discourse sample, then reflects to identify weaknesses in the current coding framework and applies revision actions (modify, add, delete, hold, merge) before labeling the next sample.
Semantic similarity heuristic for interpreting Cronbach's alpha - The article proposes a heuristic for judging when Cronbach's alpha can be read as a reliability index: researchers should examine how semantically similar a test's items are.
Semantically equivalent prompt-variant benchmark construction for measuring prompt privilege - A benchmark construction methodology that instantiates prompt privilege on the MedQA benchmark: for each original MedQA question, four additional semantically equivalent prompt variants are generated representing distinct user cohorts — Low Literacy, Everyday User, Educated User, and Prompt Engineer
Semi-supervised ANN framework combining ANNs with DINA and DINO diagnostic classification models - The article proposes a framework that combines artificial neural networks with two typical theoretical diagnostic classification models, the DINA model and the DINO model, within a semi-supervised learning framework.
Semi-supervised deep learning bias-adjustment framework for IRT parameter estimation under nonignorable missingness - The article introduces a framework combining semi-supervised learning with deep learning to adjust biased 2PL-IRT parameter estimates caused by nonignorable missingness in VLE data.
Sentence-level social network analysis pipeline for eye-tracking reading data - A reusable analytic pipeline in which fixations are filtered (Tobii I-VT), sentence AOIs are classified as topic-introducing, topic-medial or topic-ending, AOI strings are converted to transition matrices, and weighted networks are built in R with igraph to compute strength, betweenness, harmonic ce
Seven primary team learning behaviours framework - The review adopts the framework of seven primary team learning behaviours: knowledge sharing, constructive conflict, co-construction, boundary crossing, team activity, team reflexivity, and storage and retrieval.
Seven-category teacher instructional move coding scheme for collaborative small-group discussions - The study adapted a coding scheme from Lin et al.
Seven-dimensional outcome comparison framework for analyzing implicit learning outcomes in summative assessments - The article develops and applies a comparison framework that identifies characteristics of summative assessments in seven dimensions: cognitive process, transferability, depth of analysis, interdependence, novelty, scaffolding and communication.
Seven-phase framework for reflecting on how AI methodologies may transform science education research - The article proposes a framework of seven epistemological phases for AI-based educational research: \"(1) problem framing, (2) instrumentation and measurement, (3) experimentation and evidence‑based inference, (4) comparisons and replication, (5) building norms and consensus, (6) implementation and i
Seven-step recursive ANT analysis of dialogic and non-dialogic classroom talk networks - Applying ANT to classroom talk, the article defines networks of human and non-human actors (texts, boards, instructions) that constitute the socio-material structure of talk.
Seven-step recursive ANT analysis procedure for classroom discourse - The article adapts ANT into a recursive analytic procedure for classroom talk with seven steps: recognizing actors (human and non-human), tracing connections into assemblages, identifying translations of roles, identifying shifts and changes in the network, delineating interactions' outcomes, examin
Seven-type feedback focus taxonomy adapted from Narciss for written teacher feedback - The article adapts and refines Narciss's content-based feedback taxonomy into seven feedback focus types for annotating written feedback: Task Constraints, Concepts, Elaboration, Clarification, Mistakes, Praise, and Other.
Simulated video group discussion method for isolating cognitive activities in collaborative learning - A laboratory research method in which participants watch a video-recorded group discussion performed by actors, so that all participants share the same prior knowledge and are exposed to the same group process.
Simulated-respondent framework - The framework bins student ability into L labeled intervals (Table 2, from Foundational to Elite mastery), prompts the LLM to answer as a student of a given ability level with the correct answer supplied, and trains it with an MSE loss on next-token option probabilities against interval-averaged MCM
Simulation video coding scheme for behaviors and SRL/SSRL strategies - The research team developed a coding scheme for (1) behaviors (posture and gestures, facial expressions, and vocalics) and (2) regulation strategies including SRL codes (C.2.1–C.2.6) and SSRL codes (C.3.1–C.3.2), plus social climate and participation codes (C.3.3–C.3.6), adapted from prior literatur
Simulation-based framework for evaluating learner task-selection strategies and system constraints in mastery-based ITS - A simulation framework that models mastery learning as a repeated problem-level decision cycle in which simulated learners select skills via predefined strategies, the system selects problems, performance is generated via AFM, and mastery is updated via BKT with a 95% mastery threshold.
Situated-interaction analysis of CSCL processes - The editorial's closing design recommendation for CSCL research: computer-supported collaborative learning opens a multi-dimensional shared world in an evolving, co-created context, where \"Knowledge, roles, space, and time are not simple givens whose characteristics can be assumed\".
Six discourse functions mapped onto inquiry-based knowledge construction categories - The article inductively derived a coding taxonomy: \"This resulted in the following six DF: reviewing directions and strategies, sharing and confirming information, brainstorming, empirical evaluation, clarification and elaboration, and submission and reflection.\" These map onto inquiry-based learnin
Six patterns of theoretical integration in EER research - The review develops a six-category typology of how deeply EER articles integrate theory, classifying each of the 24 included articles along four dimensions: didactic considerations, implementation of a specific EER, theoretical foundation connected with didactic considerations, and empirical verific
Six-behaviour spatial pedagogy coding scheme for team teaching - An observation coding scheme adapted from prior work that integrates spatial pedagogy space types with team-teaching research.
Six-code thematic framework for knowledge claims in ITE-diversity research - The review developed an inductive, iterative coding framework to classify knowledge claims about preparing teachers for diverse learners.
Six-dimension taxonomy of faculty-student relationships extracted from over a dozen measures - To handle heterogeneity across FSR measures, the authors inductively reviewed more than a dozen survey instruments and derived a categorical coding scheme.
Six-dimensional framework of authentic assessment characteristics - The review proposes a six-dimensional framework for authentic assessment (AA) characteristics consisting of authenticity in assessment, cognitive challenges, assessment criteria, feedback, student agency and social collaboration.
SMA-grid and what/when/how/by-whom-or-what frameworks for characterizing SRL scaffolds - The introduction uses two frameworks to compare the special-issue studies.
SME-elicited LLM feature-based item difficulty prediction - A catalog of item features developed through focus groups with four subject matter experts from the assessment design team and cross-referenced with prior literature, then rated by GPT-4o for each item.
Social network analysis for e-learning interaction networks - The review frames social network analysis (SNA) as a methodological approach for describing and analysing network behaviours in e-learning.
Socio-cognitive openness framework: three levels, two dimensions, four discourse acts, and eight components - A conceptual framework for analyzing openness in online knowledge building contributions, organized at three levels: dimensions, discourse acts and components (see Fig.
Spatially explicit Markov simulation of educational network growth - A discrete-time stochastic model with annual transitions representing both school robotics-program development and each school's relationship to the ARC network.
SREM - The article develops a shared random effects model (SREM) to estimate English proficiency development and time to reclassification simultaneously.
SRJudge: three-stage Select-Reason-Judge framework for fine-grained knowledge concept tagging - SRJudge decomposes knowledge concept tagging into three stages.
Standardized taxonomy descriptors for CSCL phenomena and analytic approaches - Because CSCL lacks shared descriptors for phenomena and analytic approaches, the article recommends standardized descriptors within a community taxonomy.
State–trait temporality framework for child engagement measurement - The review organizes variation in engagement measurement around assumptions about temporality: engagement can be treated as a momentary, transient state of immersion in an activity, assessed by observations and experience sampling, or as a relatively stable tendency to behave in certain ways, assess
Stealth assessment - Stealth assessment is an approach in which measurement of a skill is woven into gameplay so that students' ordinary choices generate the assessment data without them knowing they are being assessed.
Stereotype Congruence Rate (SCR) - SCR is a complementary bias metric the article introduces because existing benchmarks \"probe model-internal associations via templated inputs and cover only broad occupational categories\", which does not suit free-text career recommendations.
Stimulated recall procedure for eliciting learner thinking - The study used a stimulated recall procedure in which students' thinking processes during the tutorial group discussion were tapped, producing recall protocols analysed alongside verbatim transcripts of the group's verbal interaction.
Strategy-based temporal analytics framework for stealth assessment using LSTM evidence models - A stealth assessment framework that derives temporal evidence from students' dynamic in-game strategy use rather than only from ECD evidence modeling.
Student Digital Twin - A Student Digital Twin is defined as \"a continuously evolving model of an individual learner's academic performance, learning behavior, career interests, degree progression, financial circumstances, co-curricular engagement, and workforce preparedness.\" Its purpose is \"to simulate\": answering counte
Student Momentum Indicator - The Student Momentum Indicator is a grounded-theory metric defined as \"the percent of students enrolled in the course who finished\", calculated after each lesson as M = (total enrolled - dropouts remaining after the lesson) / total enrolled.
Subdomain taxonomy for physical activity effects: four executive function, three attention and three academic performance subdomains - The meta-analysis organises its outcome measures into a taxonomy of cognitive and academic subdomains: \"Four subdomains of executive functions (inhibition, working memory, cognitive flexibility and planning), three subdomains of attention (selective, divided and sustained) and three subdomains of ac
Subgroup responder-rate analysis for judging practical significance - The article recommends that responder rates be presented for different subgroups of students.
Subject-level demographic MAE-gap fairness regularization for multimodal attention models - An in-processing fairness framework that adds a demographic disparity penalty to the training objective of a multimodal attention model.
Suitability and feasibility framework for deciding whether to use state tests in education experiments - The paper organizes the decision of whether to use state assessment data in an RCT around two concepts from measurement theory.
Summer-dip measurement of educational ChatGPT use - A measurement strategy that infers the share of an area's ChatGPT activity tied to school from the ratio of average activity while school is in session to average activity in the summer.
SWIM: proficiency-conditioned essay generation and proficiency-alignment evaluation - SWIM formalizes student writing simulation as proficiency-conditioned essay generation: given a writing prompt and a target trait-score profile, a generator produces an essay exhibiting the specified writing behavior, and \"an AES-based evaluation framework for\" measuring profile alignment scores the
Synthetic benchmark development and stress-testing for trace-based learning-analytics evaluation - When participant access is unavailable, release a synthetic dataset mirroring the intended logging schema and study design so that trace-based measures, modelling workflows, and analysis pipelines can be developed, compared, and stress-tested openly.
Systematic review - The article defines systematic reviews as a distinct genre of research synthesis.
Systematic screening for opportunistic experiments - The guide recommends that researchers and policymakers \"discuss issues to consider when identifying potential opportunistic experiments\" before undertaking one, treating candidate screening as a deliberate step rather than an afterthought.

T¶

TACT dual-taxonomy framework for annotating tutoring moves and conditional feedback strategies - TACT organizes ESL tutoring as a pedagogical decision space connecting learner-state annotation, tutor-strategy selection, and next-turn response generation.
Taxonomy of algebraic thinking operationalizations: global measures versus specific facets - The review organizes algebra outcome measures into a taxonomy distinguishing global constructs from specific facets.
Taxonomy of causality theories: sequential, simultaneous, and reciprocal - The article organizes philosophical, theological, and scientific perspectives on causation into three families relevant to interpreting cross-lagged panel models.
Taxonomy of EAI modalities in education - The review organizes the educational AI literature by intervention type, finding that studies \"address the potential of EAI on student performance through applications (n = 8), simulations (n = 8), VR (n = 4) and AR (n = 2)\".
Taxonomy of early-grades text characteristics organized by linguistic level from sounds to text density - Table S1 organizes quantifiable text characteristics by linguistic level: sounds in words (number of phonemes, phonemic Levenshtein distance, internal phonemic predictability), word structure (decoding demand, orthographic Levenshtein distance, syllables, sight words), word meaning (age of acquisiti
Taxonomy of eight primary self-management intervention configurations - The review organizes the school-based self-management literature by the specific intervention configuration employed, distinguishing approaches such as prompted self-monitoring, self-monitoring plus reinforcement based on both accuracy and performance, and prompted self-monitoring plus graphing.
Taxonomy of five computational thinking aspects with 13 sub-aspects - From a review of 28 manuscripts defining computational thinking, the authors identified five key CT aspects in the context of modeling: decomposing problems so they are computationally solvable, creating artifacts using algorithmic thinking, generating/organizing/interpreting data, testing and debug
Taxonomy of four SR research traditions distinguished by core explanatory construct families and Lakatosian hard cores - Using a Lakatosian framework of research programs, the article categorizes SR research into four traditions, each with a hard core of foundational assertions protected by a belt of auxiliary hypotheses and methods.
Taxonomy of interaction behaviors with AI code suggestions - The article synthesizes fragmented prior characterizations of AI code completion interactions into a unified taxonomy of behavioral metrics.
Taxonomy of knowledge components classified by task features, response, relationship, and rationale - The article defines a knowledge component as 'an acquired unit of cognitive function or structure that can be inferred from performance on a set of related tasks' and classifies KCs along four dimensions: whether task features and response are constant or variable, whether the feature-response relat
Taxonomy of learning-approaches measurement inventories (SPQ, R-SPQ-2F, ASI, ALSI, ASSIST) - The review describes the family of questionnaires used to measure learning approaches: Biggs' Study Process Questionnaire (SPQ), whose strategic approach is termed the achieving approach; its revision, the R-SPQ-2F, which \"uses fewer items to narrow in on solely the deep and surface approaches\"; Ent
Taxonomy of PDW payoff schemes: four-cell matrices versus free-choice designs - The review organizes the heterogeneous payoff systems used in post-decision wagering into two broad classes: four-cell matrices, where a low and a high wager option combine with correct and incorrect responses to yield four possible outcomes, and free-choice designs with more than four cells.
Taxonomy of pedagogical sycophancy - The paper defines pedagogical sycophancy as pressure-contingent validation of a misconception after the tutor is pushed to agree, distinguishing it from ordinary factual error.
Taxonomy of reciprocal effect panel models: CLPM, RI-CLPM, PCPM, and CCLPM with lag0, lag1, and lag2 effects - The article's methodological contribution is a taxonomy of structural equation models for directional ordering, extending Muthén and Asparouhov's contemporaneous-effects models with latent measurement factors, lag2 paths, random intercepts, and covariate strategies.
Taxonomy of seven teacher humour styles and three affiliative sub-styles - The review organises teachers' classroom humour into seven styles drawn from the included literature: course-related, course-unrelated, affiliative, aggressive, self-enhancing, self-defeating, and festive humour.
Taxonomy of teacher intervention types during “making rounds” in computer-based collaborative learning - The paper organizes its analysis into distinct types of teacher intervention during rounds: ratifying pupils' work, reminding pupils of neglected task aspects, making suggestions, maintaining classroom control, and making whole-class announcements.
Taxonomy of teaching approaches used to deliver school-based RSE - The review classifies the pedagogical approaches reported in RSE studies into categories including multiple strategies, health promotion, delivery personnel, a suite of active learning pedagogies, cross-curriculum approaches, UNESCO guidance, philosophy/ethics, and sex/gender-based approaches.
Taxonomy of test-score-based school evaluation estimators - The paper organizes approaches to using student test scores for school evaluation into distinct estimator families.
TeachBench - TeachBench evaluates LLMs as teacher agents by having them conduct multi-turn instruction over structured knowledge points derived from an exam syllabus, with a fixed-capability student agent as a learner proxy.
Teacher nomination procedure for observer-reported emotion regulation - Observer reports were collected through teacher nominations in which teachers selected up to 5 students they considered at the high and low ends of emotion regulation ability from alphabetical lists of students they taught.
Teacher noticing framework: detection and interpretation of classroom situations - The article adopts the teacher noticing framework of Van Es and Sherin to model how teachers make sense of classroom and dashboard situations.
Teacher Noticing of Rationality Framework (TNRF) - The TNRF builds on van Es and Sherin's expanded noticing framework of attending, interpreting, and shaping, and \"extends van Es and Sherin's (2021) framework by specifying the object of noticing more precisely as students' mathematical thinking\".
Teacher Team Method for co-taught value-added estimation - The Teacher Team Method is one of three methods the article proposes for estimating teacher value added when students are co-taught.
Teacher value-added modeling for performance classification - The study defines teacher performance using value-added estimates from a student-level growth model with teacher fixed effects.
Teaching Through Interactions framework - The review adopts the Teaching Through Interactions (TTI) framework as its organizing model for classroom-level interaction quality, integrating attachment theory and self-determination theory.
Teaching-over-solving diagnostic - The article proposes a lightweight diagnostic that reaggregates public tutoring benchmark scores into two composites: a solving composite (problem solving, solution correctness, mistake location, mistake correction) and a pedagogy composite (socratic questioning, scaffolding win rate, pedagogy instr
Teaching-system episode framework: knowing–teaching–acting across three capability surfaces - The article proposes that tutor-agent readiness requires three separable capability surfaces: Stage 1 pedagogical judgment (a bounded teacher decision from packaged evidence), Stage 2 situated tutoring (a history-dependent adaptive teaching policy over a multi-turn episode), and Stage 3 end-to-end L
Temporal network analysis - Temporal networks are a method that models both the relational aspects (who interacts with whom) and the temporal aspects (order, duration, concurrency) of collaborative learning.
Temporal Smoothness Doubly Robust (TSDR) framework for debiased knowledge tracing - TSDR is a model-agnostic debiasing framework for knowledge tracing that integrates a doubly robust estimator — combining Inverse Propensity Weighting with error imputation — with a temporal smoothness regularizer on the imputation model's latent trajectory.
Test disengagement adjustment through sample removal and rapid-guess score correction - The article examines two techniques for accounting for test disengagement: \"removing unengaged test takers from the sample\" and \"adjusting test scores to remove rapidly guessed items\".
Testlet-level CFA for assessing dimensionality of CAT data - Because CAT administrations give different students different items, item-level factor analysis is impractical: for MAP, \"typical missing rates are around 98%\".
Text-based cognitive mapping method for measuring team shared mental models - The study measures shared mental models using Carley's (1997) automated text-based cognitive mapping: team members' open-ended answers are coded as concept networks via filtering (delete lists and a thesaurus) and windowing (relations within a range of eight concepts), and the team's shared model is
The effort-moderated IRT model of test-taking motivation - The effort-moderated IRT model (S.
The LoGeT model: localize, generalize, and transfer framework - The LoGeT model is a methodological framework for generating educational evidence that is localized, generalizable, and transferable.
The Socratic Test framework - The Socratic Test is an adaptive, typed conversational assessment that standardizes the method of measurement without standardizing prompts.
The Statistical–Mechanism Gap Fallacy: statistical contextual indicators are not psychologically transparent mechanisms - The article names and defines the Statistical–Mechanism Gap Fallacy as \"the tendency to collapse the gap between statistical indicators and the psychological, social, or cognitive mechanisms they are presumed to represent\".
Thematic unit analysis for multiple-text writing integration - The study introduces the thematic unit, defined as \"an idea or a chain of connected ideas\", as a new unit of analysis for examining integration in written products.
Theoretically productive classroom dialogue framework: five recurring themes plus the cross-cutting theme of high student contribution - The article defines theoretically productive dialogue as classroom verbal interaction complying with five recurring themes: open questions, elaboration of previous contributions, reasoned discussion of competing viewpoints, linkage and coordination across contributions, and meta-cognitive engagement
Theoretically refined multidimensional interaction rating scheme - The editorial reports that Meier, Spada and Rummel differentiated as many as nine dimensions of interaction for quantitative analysis, derived through bottom-up qualitative content analysis refined by top-down theory-informed considerations.
Theory- or empirical-evidence-grounded responder-analysis cutoff selection - The article acknowledges the criticism that cutoffs for classifying students' percentage change into categories may seem arbitrary, and answers it: \"the cutoffs may come from theory; for example, from expert considerations of what should at least be learned during a course/period of time.
Theory-driven classification of six IVR learning processes and nine eye-trackable constructs - The article develops a classification mapping eye-tracking metrics onto learning processes in IVR, grounded in CTML, CATLM, CASTLE, CAMIL, and Cognitive Load Theory.
Theory-driven taxonomy of six motivation constructs for pedagogical agent research - The article organizes PA motivation research around four prominent theories — social cognitive theory, situated expectancy-value theory, interest theory, and self-determination theory — arguing that lumping motivation into one meta-construct is \"theoretically unstable.\" From these theories it derive
Theory-driven units of analysis for educational dialogue across multiple timescales - The Baker, Schwarz, and Ludvigsen Squib argues for deep theoretical understanding of discussion and its unit of analysis to inform methods for analyzing dialogue over different timescales, from episodes to societal-scale argumentative knowledge construction.
Thick description as the analytic basis for understanding transactive discourse - Stahl draws on Austin and Geertz's notion of thick description: CSCL researchers understand transaction meanings because they share a community with their subjects.
Think-aloud protocol validity framework: reactivity and veridicality as the two central methodological concerns - The review organizes methodological concerns about think-aloud protocols around two constructs.
Thinking-aloud protocol scoring system for math metacognitive skillfulness - An observational coding system in which thinking-aloud protocols are analyzed for spontaneous metacognitive activities during math problem solving.
Threaded analysis-unit coding for small-group chat communication - A design principle for chat communication analysis methodology: coding procedures for small group chat should build in threading, i.e.
Three accountability methods for setting English Learners' academic progress and proficiency expectations - Chapter IV presents an organizing framework of methods for setting academic expectations for English Learners.
Three assumptions for causal interpretation of cross-lagged panel models: consistency, strong ignorability, and positivity - Drawing on the Rubin causal model and potential outcomes, the article specifies the assumptions under which cross-lagged coefficients may be read causally.
Three categories of attentional effects (canonical, phase-crossover, attention-crossover) for discriminating process models - The article proposes a categorization of effects on cognitive processing during word-problem solving: \"Canonical effects\" are predicted by all process models; \"Phase-crossover effects\" occur when a difficulty or ability affects the phase opposite to its canonical one (predicted by flexible models);
Three complementary views of AI-integrated writing: product, process, and interaction - The article argues that understanding student writing in AI-mediated settings requires three complementary views: the product view characterizes the final artifact, the process view captures how the artifact develops over time, and the interaction view captures how the student engages with the AI du
Three interacting themes for examining CSCL classroom innovation - The article proposes a conceptual framework of three interacting themes for examining how CSCL research creates innovation in classrooms: \"context and systemic change, capacity and community building, and innovation as a process of inquiry.\" These themes organize the analysis of knowledge-building i
Three levels of interpretational aid in CSCL teacher dashboards: mirroring, alerting, and advising - Building on Soller et al.'s algorithm for collaboration support tools, the article distinguishes dashboards by which agent - tool or teacher - performs detection and interpretation.
Three measures of test adaptation based on optimal test information - The article introduces a framework of three new measures of test adaptation, each grounded in the amount of information that a test provides to individual examinees.
Three multidimensional IRT models for testlet-based tests - The report describes three multidimensional IRT models that account for conditional dependence among items sharing a testlet stimulus by incorporating testlet-specific dimensions.
Three paradigms of AI integration in education: AI-directed, AI-supported, and AI-empowered - The review adopts Ouyang and Jiao's (2021) framework, which \"frames the educational role of AI along three paradigms that reflect increasing levels of learner agency\": AI-directed (learner-as-receiver, rooted in behaviorism, with AI structuring content and giving prescriptive feedback), AI-supported
Three Paths of Student-AI Interaction typology - A typological framework identifying three qualitatively distinct modes of student-AI engagement: Passive Review (student accepts AI output with minimal scrutiny, unidirectional, limited higher-order opportunity), Direct Question (transactional gap-filling with surface evaluation, moderate opportunit
Three planes of learning in CSCL: individual, small-group, and community - The article names a tripartite unit-of-analysis framework for CSCL: \"analyses of CSCL are often conducted on one of three levels: individual learning, small-group cognition or community knowledge building\".
Three pooling formulations for Bayesian Knowledge Tracing (standard, multi, and hierarchical) - StanBKT organizes BKT formulations into three pooling types.
Three roles of AI in generative learning: Generation Partner, Generation Substitute, Generation Catalyst - A typology of three roles AI can occupy relative to learners' generative processes.
Three routes to theoretical integration: competition among theories, unified frameworks, and bridge-building integration of existing theories - Drawing on Gigerenzer (2010), the essay presents three routes researchers can take toward integration.
Three structural causal models of the self-concept–achievement reciprocal relationship - The article derives a taxonomy of three structural causal models for the reciprocal effects model: \"(a) a model with prior scores predicting post scores; (b) a model with additional lag-2 effects ...
Three validation targets for educational simulation: behavioral believability, educational mechanism fidelity, and institutional counterfactual usefulness - The paper argues that educational simulators are too often evaluated only on surface realism, and distinguishes three targets that are often conflated.
Three-category taxonomy of collaborative instruction models with distinct causal mechanisms - The review organizes two-teacher in-class interventions into three categories studied largely separately: co-teaching (general plus special education professionals), teacher assistants/aides (collaboration with adults without formal teaching qualifications), and team teaching (two general education
Three-category taxonomy of educational research evidence - The article organizes science education evidence into three research categories with different methodological strengths and different implications for policy.
Three-component decomposition of teacher grade effects - The article conceptualizes teacher influence on student grades in three ways: total effects on ninth-grade course grades, persistent effects on students' grades in the same subject the following fall, and grading mismatch between the grades given and students' subsequent performance.
Three-construct framework for measuring challenge navigation in games - The article organizes in-game motivational behavior into three constructs: \"challenge approach, challenge response, and general persistence\".
Three-criterion framework for evaluating test mode comparability - The report organizes its comparability evaluation around three criteria drawn from the score-comparability literature: \"there are three general categories of criteria adopted in evaluating comparability: (1) content and construct validity, (2) psychometric properties and reliability, and (3) statist
Three-design-decision framework for spatial ability research in mathematics education - The review proposes a guiding framework identifying three major design decisions for scholars using spatial skills in mathematics education research: selecting a spatial ability taxonomy, selecting corresponding analytical frameworks, and selecting spatial tasks for assessing spatial performance.
Three-dimensional coding framework for monitoring events (target, valence, phase) to identify adaptive regulation and maladaptive behavior in collaborative learning - The study contributes an empirical framework, built on the socially shared regulation of learning (SSRL) framework, for identifying adaptive regulation and maladaptive behavior from process data.
Three-dimensional parental scaffolding framework - The article organises parental scaffolding during joint problem-solving into three aspects: \"cognitive support, emotional support, and contingency\".
Three-dimensional taxonomy for organizing research on primary school students' creativity - The article introduces a three-dimensional taxonomy as the organizing scheme for its empirical integration.
Three-family taxonomy of spatial ability structures: specific-factor, broad-factor, and unitary - The review organizes the sprawling spatial ability literature into three general types of taxonomies.
Three-group coding framework for scientific habits of mind in informal game forums - The article develops a set of 18 codes (following Chi 1997) based on AAAS benchmarks for scientific literacy, Chinn and Malhotra's framework for evaluating inquiry tasks, and Kuhn's epistemological framework.
Three-layer hierarchical framework: Pellegrino's assessment triangle, Webb's Depth of Knowledge, and the TIMSS cognitive domains - The article organizes its analysis through three complementary frameworks in a hierarchical relationship: Pellegrino et al.'s assessment triangle (cognition, observation, interpretation) as meta-theory, Webb's Depth of Knowledge as taxonomic theory, and the TIMSS framework as the operational layer.
Three-layered AI-driven learning analytics framework for examining collaborative problem solving - The article proposes a three-layered analytical framework integrating AI algorithms with learning analytics to examine collaborative problem solving as complex adaptive systems: \"AI algorithms were incorporated into that framework using multiple learning analytic approaches\".
Three-level engagement typology: information sharing, feedback, and collaboration - Adapting Cashman et al.'s four levels of interaction, the report categorizes Core to College network engagement into three primary kinds: \"Information Sharing—providing information and/or raising awareness about the Common Core State Standards and aligned assessments\"; \"Feedback—gathering feedback a
Three-method framework for accounting for co-teaching in value-added models - The article organizes the problem of modeling value added \"when the same student is taught the same subject by multiple teachers\" into three named methods: the Partial Credit Method, the Teacher Team Method, and the Full Roster Method.
Three-part taxonomy of methods for studying variation in treatment effects - The report organizes the quantitative methods literature into three research topics: subgroup (moderator) analyses using participants' characteristics \"measured before the intervention is implemented\"; subgroup analyses based on \"participants' experiences, mediators, and outcomes measured after prog
Three-question framework for reviewing noncognitive measurement and segmentation in postsecondary education - The report organizes its targeted document review and expert interviews around three questions relevant to the Student Segmentation Initiative: what instruments and measures are available to assess postsecondary students' noncognitive attributes; to what extent these instruments are used to classify
Three-question structure for school performance inquiry - The framework structures school performance inquiry around three questions: \"1.
Three-question taxonomy of motivation constructs: beliefs, task value, and reasons for engaging - The article organizes motivation constructs using Graham and Weiner's three basic questions — \"Can I do it?\", \"Do I want to do it?\" and \"Why am I doing this task?\" — mapping them onto competence and control beliefs, feelings-related and importance-related task value, and intrinsic versus extrinsic r
Three-step framework for generating validated natural-language explanations of question difficulty - The article introduces a framework that treats difficulty as \"a quantity to be explained\" rather than only estimated.
Three-tiered classification system of self-regulation: categories, domains, and facets - The review establishes a hierarchical taxonomy for organizing the variety of self-regulation terms found in early-childhood measures.
Tied-duration analysis of uptake and relevance in collaborative discourse - Tied duration is the total time an interaction process remains actively receiving responses from the group, computed from the temporal network by aggregating active intervals from first to last uptake, after trimming idle periods.
Time-course and mixed-effects analysis of educational eye-tracking data - The author argues that eye tracking's main advantage is detailed time-course information, which total fixation time or gaze duration fails to exploit because it \"sums up all fixations or gazes made to a particular area of interest.\" She recommends first-pass and reinspection measures, fixation proba
Tool-supported workflow for identifying, clustering, and visualizing common modelling mistakes as bug fixing transformations - A three-step workflow in which candidates for common mistakes are represented as bug fixing transformations that locally rewrite incorrect syntax trees into correct ones.
TRACE (TRansformer for Academic Course-grade Estimation) - TRACE (a TRansformer for Academic Course-grade Estimation) encodes a student's major, course and grade history with separate embedding layers and positional encodings based on relative semester, so all courses in the same semester share a temporal vector, enforcing a permutation-invariant representa
Trace parser with action library and process library for measuring SRL from multi-channel data - A trace-based measurement protocol that converts raw multi-channel trace data into SRL processes.
Trajectory-based early indicator framework for college readiness monitoring - The article presents a framework that applies college readiness benchmarks to longitudinal middle-grades assessment data, assigning each student a series of up to six on-track status indicators and grouping students into six trajectory patterns.
Transactivity coding scheme for self- and other-oriented reasoning - The article operationalizes transactivity as reasoning made explicit that builds on prior reasoning in a discussion.
Transcript-based interactional fluency and direction-sensitive uptake analysis - A reusable analysis pipeline that operationalizes interactional fluency and direction-sensitive linguistic uptake for spoken AI practice \"using metrics extractable from diarized ASR output without manual linguistic annotation.\" The fluency component extracts seven features from word-level timestamps
Transformation-and-re-estimation method for probing scale sensitivity of teacher value added - The article introduces an approach for testing whether teacher value added depends on the test scale: estimate value added on the original scale, then \"applying extremely mild nonlinear transformations to the original scale and re-estimating the value added.\" Because \"by definition at most one of th
Transformer-encoder knowledge tracing framework with IEU/IIU input vectors - A deep knowledge tracing framework for the ASSISTments dataset that predicts students' responses to end-of-unit test problems from action logs of in-unit assignments.
Transition Network Analysis (TNA) - Transition Network Analysis (TNA) models learning processes as networks of discrete behavioural states connected by directed, probabilistic transitions, computed here with a web-based TNA Shiny app.
Triangulated multi-source assessment of self-regulated reading processes - Because think-aloud protocols face reactivity and veridicality risks, the review recommends that researchers \"view think-aloud data as indicators of only a part of self-regulated reading process rather than a full picture of it\".
Triangulated multi-source assessment of writing processes and writing knowledge - A methodological framework combining three elements to examine writing processes of struggling adult writers: observation of collaborative writing tasks, analyses of written texts, and structured individual interviews with retrospective and prospective parts.
Triangulated three-method analysis of collaborative problem-solving enactment - The article develops a triangulated analysis method combining three equally valued approaches to understand CPS at group level: \"(a) directed content analysis of the actualized CPS in groups, (b) process analysis and visualizations, and (c) qualitative cases\".
trifactor model - The trifactor model is a structural equation modeling approach developed by Bauer et al.
Trio-ethnography - Trio-ethnography brings three participants (here two computing educators with contrasting instructional philosophies and one undergraduate student) into reflective conversation as co-inquirers.
Trustworthiness lens for evaluating LLMs in education - The article proposes adapting trustworthiness criteria from the ML community as a structured lens for LLM evaluation in education.
Turn-level and lesson-level classroom dialogue coding - An observational scheme coding every speaker turn for codes such as ELI, EL, REI, RE, OI, UC, Q, SC, RC, RB and RW, plus five holistic lesson-level ratings (Aims, Monitoring, Reflection, Talk Rules, Student Participation).
Tutoring Effectiveness Index (TEI) - TEI is a fixed-weight linear index over four signals computed without any external model at selection time: \"a Schoenfeld-Verify keyword ratio, a math-step density, an ends-question rate, and a deep-reasoning gate from the Deep-Thinking Ratio (DTR) probe\".
Two complementary misconception-acquisition formalizations: the Novice Student Misconception Model and the Expert Tutor Misconception Model - The article formalizes misconception acquisition from two perspectives.
Two goal-relative criteria — specification and assessment demand — locate operations on a graded peripheral–constitutive continuum - The framework supplies criteria applicable before instruction for classifying cognitive operations relative to a learning goal.
Two views on temporality in self- and socially regulated learning research - The editorial proposes that existing research on the temporal characteristics of S-SRL can be viewed through two distinct perspectives.
Two-axis expression-and-activity taxonomy of scaffolding behaviours - Beyond cognitive strategy, the review codes scaffolding behaviours by method of expression (verbal/auditory vs.
Two-dimensional classification of diagnostic situations: information base and collaboration requirement - Adopting Heitzmann et al.
Two-dimensional conceptual framework for student–GenAI interaction and perception in STEAM learning - The article proposes a conceptual framework with two dimensions: student–GenAI interaction and students' perceptions of GenAI use.
Two-dimensional framework for classifying instructional support in peer-feedback - The article proposes two orthogonal ways to differentiate instructional support in peer-feedback scenarios.
Two-dimensional rubric coding operational process category and depth of explicit metacognitive evidence - The article introduces a shared coding rubric applied identically to interaction logs and think-aloud speech.
Two-dimensional taxonomy of embodied learning environments for graphing motion based on bodily involvement and immediacy - The review proposes categorizing embodied learning environments for graphing motion on two dimensions: bodily involvement (own motion vs.
Two-dimensional typology of eWIL: function of technology by degree of technological involvement - The review develops a typology classifying eWIL experiences along two dimensions: the function of the technology (supporting administrative WIL processes vs.
Two-form AI difficulty ruler (solving-based and review-based) - The article distinguishes two forms of the AI ruler.
Two-level classification of visualizations in drawing tasks: abstractness level and representational vs. relational drawing type - The review proposes describing visualizations in learning-by-drawing studies on two levels: the level of visual abstractness (concrete vs.
Two-metric turn-level evaluation of chatbot scaffolding and student uptake - An evaluation pipeline that scores both sides of student-chatbot interactions at the turn level.
Two-phase JOS paradigm separating intuitive solvability judgments from later solving outcomes - The two-phase paradigm separates the JOS phase, where participants make intuitive solvability judgments for an entire set of problems, from a subsequent solving phase where solving outcomes and effort regulation are measured.
Two-stage LLM misconception identification pipeline - A pipeline combining quantitative quiz analysis with two LLM stages: GPT-4 (gpt-4o-2024-08-06, temperature 0.3) generates hypothesized misconceptions from structured question representations, most common incorrect answers, and Whisper-transcribed lecture transcripts; Gemini 2.0 Flash (temperature 0)
Two-stage pooled-then-specialized student-simulation training - StudentSim overcomes sparse per-student data with a two-stage pipeline.
Two-strand taxonomy of embodied learning in language education - The review organizes the empirical literature into two strands: \"embodied learning through orchestrating embodied language learning and teaching\", and \"embodied learning in naturally occurring language learning interactions\".
Two-type taxonomy of direct behavioral and latent-construct participation indicators - The review synthesizes the varied conceptual indicators used to model individual participation into two general types.
Two-type taxonomy of emergent leadership: intellectual and organizational leadership moves - The article adapts the Li et al.

U¶

Unidimensional Rasch model scaling - All MAP tests assume a single latent achievement factor and are scaled on the unidimensional Rasch model with the RIT scale, linearly transformed from logits (RIT = logit x 10 + 200).
Unsupervised Dynamic Bayesian Network model of speech style accommodation - The model measures speech style accommodation from raw speech without labeled data or hand-crafted indexical features.

V¶

Validity, reliability, and fairness of AI outputs as a practical evaluation framework grounded in educational measurement - The article proposes a practical framework for evaluating AI-generated outputs in edtech, grounded in educational measurement concepts.
Value-added modeling - Value-added measures quantify a teacher's contribution to students' learning gains.
Van de Pol et al.'s goals-and-means scaffolding framework - The review adopts Van de Pol et al.'s (2010) framework, which distinguishes scaffolding goals from scaffolding means and specifies characteristics of scaffolding, as its analytic lens.
VARIA four-property formalisation of the AIAA integrity-via-variation hypothesis - VARIA formalises the claim that an LLM can generate assessment variant sets that are simultaneously surface-distinct, construct-equivalent, rubric-applicable and difficulty-matched.
Vector quantization of group physiological data for collaborative-learning state transitions - The study treats the collaborating group as a dynamic system whose aggregated state is a three-dimensional vector of the members' heart rates, traced over the exam session.

W¶

Waitlist-cutoff regression discontinuity design using administrative course registration data - The study introduces a novel research design that exploits discontinuities in course admissions created by waitlists, applied to administrative course registration data at a community college.
Weekly β0 structural check for cohort behavioural reorganisation - A structural summary of how a cohort divides in each week: learners are represented as points in a feature space, connected when Euclidean distance falls below a threshold ε, and β0 counts the connected components of the graph, read as \"the number of distinct behavioural groups the cohort forms that
Weighted plausible-value analysis with Rubin's rules and replicate-weight variance estimation - The article recommends that secondary analysts of PVs weight student observations to restore population representativeness, repeat each analysis separately across all PVs (usually 5 to 10), and aggregate results using Rubin's rules with replicate-weight variance estimation.
WestEd course categorization scheme for EO 1110 entry-level WC and QR course types - A four-category scheme developed by WestEd researchers to classify how CSU campuses provide additional supports to students in entry-level Written Communication and QR courses under EO 1110.
What Works Clearinghouse (WWC) design standards - The What Works Clearinghouse (WWC), established in 2002 by the U.S.
What Works Clearinghouse evidence-based standards for reviewing transition program research - The review applies a structured evidence-rating framework in which programs receive categorical effectiveness ratings such as mixed effects and potentially positive effects for specific outcome domains like employment, postsecondary education, and independent living.
Within-person versus between-person cross-lagged panel modeling (RI-CLPM versus CLPM) - A methodological framework distinguishing two ways of analyzing cross-lagged panel data.
Within-student multilevel analysis of judgment accuracy awareness - The article argues that true awareness of judgment accuracy requires analyzing whether an individual student's increase in judgment accuracy is accompanied by an increase in confidence, which demands multiple measurements per student.
Within-study comparison (WSC) methodology - A within-study comparison examines \"whether and in what context nonexperimental methods can successfully replicate the results of randomized experiments.\" The article situates its contribution in this growing literature and applies a WSC to charter school effects on student achievement, using a rand
Working classification of chatbot and language-model technologies for AI coaching - The review proposes a working taxonomy of coaching technologies, from rule-based programming (fixed rules and scripted responses) through task-oriented and ML-enhanced NLP/NLU chatbots to early LLMs and mass-market general-purpose LLMs.
Workload dynamics framework: temporal metrics of workload variation, intensity, and distribution over a term - The article proposes analyzing workload beyond static or average time-on-task by quantifying its temporal patterns.
WWC attrition model - The WWC attrition model is a framework for limiting attrition bias in assessments of intervention effectiveness: it establishes a standard for acceptable levels of sample attrition in randomized controlled trials by deriving bounds on the bias attrition can induce.
WWC effectiveness rating framework for intervention evidence - The What Works Clearinghouse rates interventions per outcome domain by combining study design quality with the magnitude, statistical significance, and consistency of findings.

Z¶

Zirbes's periodical-search method for identifying educational research problems - The article describes a search procedure that reads recent issues of nine educational periodicals and extracts statements of unsolved research problems, which are usually mentioned incidentally in connection with practical situations.