Skip to content

Epistemic Cognition

Principle · All principles
Evidence · 28 claims (13 for, 13 mixed, 2 against) · 39 studies (16 causal, 6 theoretical, 5 review, 4 quant-synthesis, 4 qualitative, 3 associational, 1 design), q1–q4 · 4 of 39 report an effect size · 20 claims rest on one study

Conditional relationship

Epistemic cognition is reasoning about knowledge itself: what counts as evidence for a claim, how a claim is justified, how sure to be, and when to revise. As a design principle it says: when a learner has to judge claims rather than only recall them, teach the standards for judging that kind of claim explicitly, have the learner apply them to real claims and justify each judgment, and check the judgment on new claims later.

For whom and on what. The evidence recorded here concerns adolescent and adult learners (middle-school science students, high-school government students, eleventh graders, first-year university students in a critical-thinking course, in psychology and in a library session) and two groups of child learners who were observed, not taught (grades 2–6 judging explanations; K-8 pupils trying out a chatbot). The knowledge targeted is a procedure (moves for checking an unfamiliar web source) or a principle (the ideals by which a discipline judges evidence, such as valid methods and replication), and in argumentation a complex-skill (writing or arguing a case with evidence). The starting response that calls for this principle is a judgment made on the wrong grounds: trusting a page because of its logo or domain, judging a study by whether it matches one's own experience, or offering an opinion where a reason is asked for.

What is expected to change, and when. Teaching these standards is associated with better judgments of the same kind of evidence, at an immediate posttest and, in one small randomised pilot, five weeks later (near-transfer to new sites or new study reports). Dialogic argumentation over several years was followed by better individual written arguments on new topics. No claim here reports a horizon beyond five weeks for evaluation skills, and none measures whether learners judge evidence better in their own lives.

The limit that matters most for design. The gains are specific to the standard taught. A semester of critiquing flawed psychology studies changed how students judged a study's methods and did not change how they judged a source's credibility. And a single library session, whether it taught a checklist or a lateral-reading framework, produced modest gains with no difference between the two. So the model predicts improvement in the kind of judgment that is taught and practised, with enough sessions; it does not predict a general rise in "critical thinking" or a large change from one session. One theoretical argument qualifies this: some standards of reason assessment (not begging the question, other principles of informal logic) are subject-neutral, while others need domain knowledge (Principles of reason assessment are both subject-neutral and subject-specific [~W]). If so, a general standard taught explicitly might carry further than a domain-specific one; no claim in the wiki tests that.

No claim in the wiki tests "epistemic cognition instruction" as a whole, or measures epistemic beliefs (how certain or simple a learner thinks knowledge is) as an outcome of teaching. The claims below test particular forms of it: source evaluation, disciplinary evidence critique and argumentation.

Sibling. Inquiry-based learning models learning by investigating a question, and its default design ends with a synthesis "with explicit standards of evidence". This page owns those standards as a target in their own right, in any domain, including where no investigation happens (judging a news site, a popular report of a study, an argument). Inside an inquiry, this page's model applies to the synthesis and evaluation steps; the inquiry page's guidance on how much to guide the investigation is not repeated here. Evaluating Sources is a narrower application (credibility of sources) that has not yet been converted to this format.

Default design, while the relationship is untested

The page's earlier guidance, made concrete as a default a designer can act on and revise. Each step is a proposal; the evidence for it is stated beside it, and no claim here tests the sequence as a whole.

  1. Anchor the work in real claims the learner must judge, not in discussion about knowledge in general: a web page, a news report of a study, a published argument, two competing explanations. Untested; from the earlier page ("abstract discussion about knowledge can stay detached from practice"). Every study below used real or realistic material: live websites, popular reports of psychological findings, flawed studies to critique.
  2. Name the kind of claim and its standard before teaching. Decide which judgment is the target: who is behind a source and what others say about it; whether a study measured what it claims and has been replicated; whether an argument's evidence supports its conclusion. Teach that standard explicitly. Evidence: the critique-and-redesign course changed only the disciplinary ideals it emphasised, not lay credibility judgments (Collaborative critique-and-redesign [~M], below). Two theoretical papers from one 1988 conference argue the same from philosophy: critical thinking has to carry the methodological criteria of the discipline (Weinstein argues critical thinking needs a disciplinary perspective [+W]), and science teaching should make the philosophy of science and the comparison of rival theories explicit (Critical science education requires explicit attention to philosophy of science and rival theories [+W]). Both are arguments, not tests, and share one source.
  3. Model the expert move, then have learners do it on new cases. For web sources, show a reader leaving the page to search what independent sites say about it (lateral reading), rather than inspecting the page's own features. Dose in the studies: two 75-minute lessons (university), six 50-minute lessons (high school), eight lessons (eleventh grade). Evidence: Civic Online Reasoning Instruction Improves Evaluation [+M]. Plan more than one session: one-session instruction of either kind produced only modest gains (Checklist-based evaluation [~M]).
  4. Contrast stronger and weaker justifications. Put a well-supported and a poorly supported claim, source or study side by side and ask which is better warranted and why; in a discipline, have small groups critique a flawed study and redesign it. The critique-and-redesign course used three such tasks across a 12-week course. In science inquiry, critiquing problematic inquiry cases before running one's own investigation in a virtual lab did more for middle-school students' scientific literacy than watching a teacher demonstrate (50 students; Critiquing problematic inquiry cases before virtual-lab inquiry [+W]). So critique of a flawed case can come before learners do the work themselves, not only after. Partly supported by those two studies, one single-group and one small two-group comparison with no effect size printed; the side-by-side contrast of a strong and a weak case is untested, from the earlier page.
  5. Require a justification for every judgment: the claim, the evidence, and why that evidence warrants it, then a reply to a counterargument from a peer. Evidence: argumentation activities improved argument quality and attention to evidence (Argumentation Improves Reasoning [+M]); the Justification and Argumentation elements describe the activity.
  6. Ask for confidence, then show the evidence. With each judgment, have the learner rate how sure they are; after feedback, discuss the judgments made with high confidence that were wrong. Untested for epistemic judgment; from the earlier page ("certainty calibration is hard", "overconfident mistakes can become useful if followed by evidence-based correction"). The nearest evidence concerns memory for facts: errors made with high confidence and then corrected are remembered better than low-confidence ones, provided the feedback is immediate and clear (High-confidence errors lead to better retention after correction [~M]; a review and one experiment, no effect size recorded). That supports spending discussion time on confident errors; it does not show that confidence ratings improve the judgment standard itself.
  7. Judge the design on new claims of the same kind, at a delay, scoring the reasons as well as the verdict, and scoring separately any kind of judgment the instruction did not teach. The randomised pilot tested five weeks after the lessons; a constructed-response task was used in the eleventh-grade study. Untested as a whole; from the studies' designs.

Constraints from the earlier page, still untested: without explicit standards, discussion defaults to opinion-sharing; learners may swing between overconfidence and blanket relativism ("it's all just opinion").

Fitting the design to a situation

The three facts that most change the decision.

  • Which kind of evidence the learner must judge. Gains have been recorded only for the standard taught, so a design for web sources and a design for research studies are different designs. If unstated, ask: what will the learner have in front of them when they have to decide what to believe, and what decision follows?
  • Time available. The recorded gains come from two to eight lessons or a semester; a single session produced modest, method-independent gains. If unstated, ask: how many sessions, and is there a later point where the learner judges new material?
  • What the learner already knows about the topic and the domain. Children applied an accuracy criterion from their own biological knowledge; checking a source laterally requires knowing which outside sources are reliable. If unstated, ask: can the learner already tell a reliable reference source in this area from an unreliable one?
If the brief says… Then change… Basis
child learners (primary school) Frame questions as "how does it happen?" and have learners compare a mechanistic and a non-mechanistic explanation of a familiar phenomenon, giving a reason; keep content to what they already know, since they reject explanations that conflict with their knowledge. Do not expect web-source evaluation routines to transfer down. Elementary students prefer mechanistic explanations [+W]: grades 2–6, observed in interviews, no instruction tested
adolescent learners, science inquiry (a lab, a simulation or a virtual lab) Before learners run their own investigation, have them critique one or two problematic inquiry cases (an uncontrolled variable, a conclusion the data do not support) and say how to fix them, rather than only watching the teacher demonstrate a correct procedure. Critiquing problematic inquiry cases before virtual-lab inquiry [+W]: 50 middle-school students, critique against teacher demonstration, no effect size printed
adolescent learners, web or news sources, a school course Embed six 50-minute lessons on lateral reading in an existing subject (government, history), with teacher preparation; practise on live sites. Civic Online Reasoning Instruction Improves Evaluation [+M]: high-school field study, matched controls
adult learners, university, judging research Run small-group critique and redesign of flawed studies in the discipline across the course (three tasks in a 12-week course), and teach credibility of sources separately if it is also a goal. Collaborative critique-and-redesign [+M]: first-year psychology, single group, no control
Goal is a general disposition ("critical thinking", "scepticism") Name the specific judgments it should show up in and teach and assess each; do not assume one transfers to another. Collaborative critique-and-redesign [~M]: no change on the untaught lay-credibility judgments
Goal is writing or arguing a case with evidence Use repeated peer argument on contested questions, followed by individual essays on new topics; plan for many sessions, not one. Argumentation Improves Reasoning [+M]: multi-year study with young adolescents; online-science synthesis
Only one session (e.g. a library visit) Teach one move (leave the page and search what others say about the source) rather than a long checklist, and expect modest gains; schedule a follow-up task if at all possible. Checklist-based evaluation [~M]: one-session checklist and networked framework did not differ; gains modest
Learners will judge answers from a chatbot or generative AI Treat the AI's output as a source to check, not an authority: teach learners to cross-check it against an independent reference they trust (course notes, a worked example, a reference work), to ask follow-up questions that expose inconsistency, and to treat verification as repeated filtering rather than one accept-or-reject decision. Show children explicitly why a chatbot that answers questions they already know is not thereby reliable on questions they do not. Assess on AI answers that contain planted errors. Students describe concrete verification behaviors for GenAI explanations [+W]: ten computing students, interviews and think-aloud, describes what they did and teaches nothing; Children use a 'trust test' strategy [~W]: exploratory K-8 study, the strategy reinforced trust after correct answers. No claim in the wiki tests instruction in judging AI output
Goal is judging arguments built on data or statistics Add to the study-method standard a question about how the data were made: who was counted and who left out, by whom, for what purpose; have learners name what the data could not show. Data construction embeds privilege in data [+W]: one theoretical philosophy-of-education article; untested with learners
online-self-paced, no teacher present Provide recorded demonstrations of the expert move, live practice items with model reasoning shown after the learner commits a judgment and a confidence rating, and a delayed check on new items. Untested proposal; the online-argumentation synthesis does not report teacher presence
workplace-clinical learners (e.g. nurses judging health information) Use the claims they will meet at work (health websites, research summaries); teach both source checking and study-method judgments if both occur, and assess each. Untested proposal; one nursing study is single-session and is read in the checklist row
High-stakes assessment of the skill Assess with constructed responses on unseen claims, scoring the reason and the verdict, at a delay; a multiple-choice verdict alone can be right for the wrong reason. Untested proposal; constructed-response and five-week tests appear in the lateral-reading studies
No open web access, locked devices, or offline exam Lateral reading cannot be practised as such; supply a curated packet of outside documents about each source and teach the learner to look for them, or shift the target to judging the reasoning inside the source. Untested proposal; the lateral-reading claim's discussion names the constraint, no study tests it

Seven rows rest on a tested comparison or a study of learners (the child, both adolescent, university, general-disposition, argumentation and one-session rows), each with the population limits stated. The AI row rests on two descriptive studies that taught nothing, the data row on one theoretical argument, and the four others are untested proposals.

Observation, state and explanation

What can be observed is a judgment and the reason given for it: which source or explanation the learner trusts, what they say supports it, how sure they say they are, and what they check before deciding. "Has sophisticated epistemic beliefs", "is a critical thinker" or "understands objectivity" are inferred states. A correct verdict does not show a sound standard (it can be reached by agreeing with the learner's prior view), and a learner who can recite the standard may not apply it to a new case.

Response Plausible explanations Observation that could change the interpretation
Trusts a page because it looks professional, has a logo or a familiar domain Has no other strategy; knows to check elsewhere but is short of time or search skill; does not know which outside sources are reliable Ask the learner to check one source with a search allowed and time given; if they still stay on the page, teach the move; if they search but consult unreliable sites, teach reference sources.
"It's all just opinion; nobody can really know" Blanket relativism as a belief about knowledge; a way of avoiding a contested or identity-laden topic; lack of the domain knowledge needed to weigh the evidence Give two claims on a low-stakes topic with clearly unequal evidence; if the learner now ranks them with reasons, the response was topic-specific, not a general belief.
Correct verdict on a study, justified by "it matches my experience" Personal coherence used as the standard; knows the disciplinary standard but defaults to the familiar one; the item did not require the standard Give a study whose finding conflicts with common experience but whose method is sound; record which reason the learner gives.
Confident and wrong Over-reliance on surface cues; a misconception in the domain; overconfidence that would also show on other tasks Ask for confidence ratings across several items; check whether errors cluster on one cue (design, authority) or one topic.
Trusts a chatbot or AI answer because it got questions right that the learner already knew A "trust test" that generalises from known to unknown questions; no independent reference to check against; the AI's fluency read as authority Give an AI answer with a planted error on something the learner does not know but can look up; record whether they check it, and against what. Descriptive basis: Children use a 'trust test' strategy [~W]
Applies disciplinary critique well, judges source credibility poorly (or the reverse) Only the taught standard was learned; the two require different knowledge; the assessment for one was harder Assess both on matched material at the same time; treat them as separate targets unless both rise together.

These rows are proposals, untested with learners. A probe can shift confidence among explanations; it does not identify a cause, and each probe is itself practice in judging evidence.

Evidence and qualifications

  • Civic Online Reasoning Instruction Improves Evaluation [+M]: three studies of teaching fact-checkers' strategies. A district-wide high-school field study (n = 499; 271 taught, 228 matched controls; teacher professional development, then six 50-minute lessons in a government course) found treatment students grew significantly more in judging the credibility of digital content; the abstract calls the design matched-control, the authors' project page cluster-randomised, and it is coded q2. A randomised pilot (four university course sections, n = 67; two 75-minute lessons) found treatment students significantly more likely to gain from a pretest six weeks before to a posttest five weeks after. A single-group study of 68 eleventh graders (eight lessons) improved on three of four constructed-response tasks, with no control. No effect sizes are reported. Not yet checked against its sources. The lateral-reading claim has been merged into Civic Online Reasoning Instruction Improves Evaluation [+M] (its old slug is now an alias), so the two are one claim resting on these same three studies, not independent support. It predicts better judgments of unfamiliar web sources after several lessons; it does not predict durability beyond five weeks, changed browsing or sharing outside class, or benefit for evaluating a primary source that has no outside coverage.
  • One-session checklist (CRAAP) instruction in source evaluation did no better than the alternative methods it was compared with in two quasi-experiments [~M]: qualifies the relationship. A think-aloud study found historians and undergraduates who stayed on a site were misled by logos and domain names while fact checkers who left it judged better (descriptive; taught nothing). In a one-session library comparison with first-year business students, a CRAAP checklist and six question words did not differ on quizzes; end-of-semester papers from the question-word sections scored higher on a rubric (2.81 against 2.28 of 4), possibly because of prior library instruction. In one 110-minute session with first-term nursing students, accuracy in the checklist group rose from 41% to 59% five weeks later and did not differ from a networked framework at any time point; there was no untrained control. The claim page's own discussion says its title overstates this: the defensible reading is that one-session instruction does no better than its alternatives. Partly checked: 1 of 3 entries passes. It is marked ~ because it limits what one session of any method achieves.
  • Collaborative critique-and-redesign instruction shifts students from a value-free view of objectivity toward a procedural one [+M]: 120 first-year psychology undergraduates (24 groups of 5) at one university, a 12-week course with three small-group tasks critiquing and redesigning flawed studies, parallel written evaluations before and after. Appeal to valid methods rose from 2% to 38% of responses, to replication from 24% to 55%; naive appeal to objectivity fell from 41% to 17%, and reliance on personal experience from 69% to 52%. None of the lay-evaluation ideals (source expertise, outlet credibility) changed significantly. Single group, no control (withheld on ethical grounds), so maturation and other coursework are not ruled out. Not yet checked against its sources. It supports teaching disciplinary evidence standards and limits the relationship to the standards taught.
  • Argumentation Improves Reasoning [+M]: a meta-analysis of 46 studies (5,415 students) of online scientific argumentation reports g = 0.603 on argumentation skill and g = 0.596 on content learning, larger in K-12 than higher education (read from the abstract; comparators and study quality not checked). A multi-year quasi-experiment in which two cohorts of young adolescents argued about social issues in peer dialogues found better argument quality and more attention to evidence in individual essays on new topics than comparison groups given essay practice and whole-class discussion. Partly checked: 1 of 2 entries passes. It predicts better justification with evidence after sustained argumentation; it does not say how much is needed or how long gains last.
  • Elementary students prefer and can evaluate mechanistic explanations using explanatory power and accuracy criteria [+W]: 53 students in grades 2–6, interviewed; across 318 evaluation occasions they accepted mechanistic explanations 71% of the time against 53% for non-mechanistic ones (p < .001), citing whether an explanation answered "how", and used their own biological knowledge to reject explanations that conflicted with it. Exploratory, no instruction or comparison group, one context. Not settled: the abstract could not confirm the entry. It shows young children already apply epistemic criteria when asked; it is evidence about a starting state, not about the effect of teaching, so it is marked W.
  • Critiquing problematic inquiry cases before virtual-lab inquiry augments the effect of virtual labs on middle school students' scientific literacy more than teacher demonstration [+W]: Liu et al. (2022), 50 middle-school students, two instructional designs before virtual-lab inquiry: critiquing problematic inquiry cases, or watching a teacher demonstration. The critique design "augmented" the virtual labs' effect on scientific literacy; no effect size is printed in the text read, and assignment and baseline comparability are not recorded on the claim page. It supports critique of flawed cases as a preparation for inquiry; it does not show which part of scientific literacy changed or for how long.
  • Students describe concrete verification behaviors for GenAI explanations, treating evaluation as an iterative filter rather than a single accept-or-reject judgment [+W] and Some children in a K-8 school study used a 'trust test' strategy, asking a chatbot questions they already knew the answers to, to evaluate it as a credible information source [~W]: two qualitative studies of learners judging AI output. Ten computing students cross-checked AI analogies against course notes and worked examples, conditioned trust on provenance and probed with follow-up questions; K-8 children tested a chatbot with questions they could already answer, and correct answers reinforced their trust. Neither taught anything. They describe starting states, one more and one less sound, and are the wiki's only evidence on epistemic judgment of AI output.

How far each may be carried: the source-evaluation studies are US high-school and university classes on the open web; the critique study is one university psychology course; the argumentation evidence is mostly science and social-issue argument with adolescents; the virtual-lab study is one small middle-school comparison; the child and AI studies observed, they did not teach. None reports outcomes beyond five weeks for evaluation skill, none reports effect sizes for the evaluation studies, and two of the core claims rest on single-group designs. Keep the kind of evidence judged, the number of sessions, the comparator and the horizon when transporting these results.

The evidence by question

Each question below links a general claim that gathers the studies behind it into one table, one row per distinct study, with each study's design and effect as recorded and no averaged verdict.

Further evidence, not yet read against this model

Re-read against the model on 2026-10-19. Of the thirteen claims listed here then, eight now sit in the model above: the hypercorrection claim (default step 6), the two Weinstein-proceedings arguments for disciplinary standards (step 2), the subject-neutral and subject-specific argument (the specificity limit), the virtual-lab critique study (step 4, a situation row and the evidence list), the two studies of learners judging AI output (a situation row, a diagnostic row and the evidence list) and the data-construction argument (a situation row). What remains is relevant or was cited by a page folded in here, but does not change the model.

Objective and learner-valued goal

Ask what the learner wants to be able to decide, and why, separately from the designer's objective. A learner may value settling a question that matters to them (is this treatment safe, is this news true, who is right in this argument) while the designer's objective is a transferable standard of judgment. Agreement, divergence and uncertainty are observations. A learner who disagrees with a conclusion the evidence supports is not thereby showing poor epistemic cognition; ask what reason they give. Choose the kind of claim, the representative material, whether outside searching is permitted, and the horizon, and score the reason as well as the verdict.

For example, a first-year nursing student may want to know whether a herbal remedy her family uses actually works, while the designer's objective is that she can judge the reliability of any health website by checking who is behind it and what independent sources say. A correct verdict on the remedy, reached because a lecturer said so, establishes neither. Let her investigate the remedy as the practice case, record her reasons and confidence, and check the designer's objective separately on unfamiliar health sites several weeks later.

What would revise this model?

The expectation should weaken if controlled comparisons of comparable learners find that several sessions of explicit standards and practice do not improve judgments of new claims of the taught kind, or that gains fade well before a few weeks. If instruction in one standard is found to improve untaught kinds of judgment as well, the specificity limit above should be relaxed rather than defended. If a single well-designed session proves as effective as several, the dose guidance should change. If learners' confidence ratings track their accuracy no better after instruction, the calibration step should be dropped or redesigned. If teaching source-checking moves for the web is found not to change how learners check AI answers, the AI row should become a design of its own rather than an extension of the source-evaluation default. Do not protect the model by calling every failure a lack of motivation or domain knowledge after the fact.

A shift in the reasons learners write, a better verdict on new material, durability over months, and better judgment in their own lives are separate claims. The evidence here reaches the first two, at horizons up to five weeks, for particular kinds of evidence. It establishes no general epistemic development, no optimal dose and no measure of epistemic beliefs as an outcome.

Claims

Examples

Illustrative

Justification — Learners are asked not only for an answer or claim, but for the evidence and reasoning that warrant it.

Source comparison routines — Students evaluate multiple sources making related claims and determine which is better supported and why.

Argument critique in discussion or writing — Learners compare strong and weak uses of evidence, confidence, and counterargument rather than only choosing a side.

Key Sources

  • Hofer, B. K., & Pintrich, P. R. (1997). The development of epistemological theories. Review of Educational Research, 67(1), 88-140. https://doi.org/10.3102/00346543067001088
  • Greene, J. A., Sandoval, W. A., & Bråten, I. (Eds.). (2016). Handbook of epistemic cognition. Routledge. doi:10.4324/9781315795225
  • Weinstein, Mark, Ed.; Oxman-Michelli, Wendy, Ed. (1989). Critical Thinking: Language and Inquiry across the Disciplines. Proceedings of the Annual Conference of the Institute for Critical Thinking (Upper Montclair, New Jersey, 1988). https://eric.ed.gov/?id=ED351305
  • Haarman, Susan. (2021). The Data Should Not Speak for Itself: Epistemic Injustice and Data as Rhetoric. Philosophical Studies in Education 52. https://www.ovpes.org/
  • Rife, T. S. (2025). Critical Communication Pedagogy at the End of the World. Journal of Communication Pedagogy. https://doi.org/10.31446/JCP.2025.1.07
  • Chinn, C. A., & Malhotra, B. A. (2002). Epistemologically authentic inquiry in schools: A theoretical framework for evaluating inquiry tasks. Science Education, 86(2), 175-218. https://doi.org/10.1002/sce.10001
  • Lombardi, D., Sinatra, G. M., & Nussbaum, E. M. (2013). Plausibility reappraisals and shifts in middle school students' climate change conceptions. Learning and Instruction, 27, 50-62. https://doi.org/10.1016/j.learninstruc.2013.03.001
  • Lombardi, D., Sinatra, G. M., & Nussbaum, E. M. (2013). Plausibility judgments in conceptual change and epistemic cognition. Educational Psychologist, 48(1), 35-56. https://doi.org/10.1080/00461520.2015.1113134
  • Liu, C.-C., Wen, C.-T., Chang, H.-Y., Chang, M.-H., Lai, P.-H., Fan Chiang, S.-H., Yang, C.-W., & Hwang, F.-K. (2022). Augmenting the effect of virtual labs with "teacher demonstration" and "student critique" instructional designs to scaffold the development of scientific literacy. Instructional Science, 50, 303-333. https://doi.org/10.1007/s11251-021-09571-4
  • Bernstein, S., & Sibia, N. (2026). Flawed but Memorable: Student Critical Reception of Interest-Personalized GenAI Analogies in Computing Education. Koli Calling '26. doi:10.48550/arxiv.2609.06095
  • Vahedian Movahed, S., & Martin, F. (2025). Ask Me Anything: Exploring children's attitudes toward an age-tailored AI-powered chatbot. arXiv. doi:10.48550/arxiv.2502.14217
  • Metcalfe, J. (2017). Learning from errors. Annual Review of Psychology, 68, 465–489. https://doi.org/10.1146/annurev-psych-010416-044022
Page metadata
Field Value
type principle
id epistemic-cognition
aliases nexus-of-justification-and-application, disciplines-as-modes-of-inquiry, datafication-as-rhetoric-metadistrust, critique-language-includes-critique-of-humanism
title Epistemic Cognition
description When learners must judge claims, explicit teaching and practice of the standards for a particular kind of evidence (checking who is behind a web source, judging a study's methods, justifying an argument with evidence) is associated with better judgments of that kind, measured up to five weeks later; the gains have not been shown to carry over to other kinds of evidence.
status review
moved_from pattern/model-evidence-link-mel-reasoning-pattern
canonical True
generated by claude/unspecified on 2026-10-19
trust unverified — no one has confirmed this page's content
sources 6 — hofer-1997, weinstein-1989, haarman-2021, rife-2025, chinn-2002, lombardi-2013

Frontmatter is the page's machine-readable record — see the Schema & Guide.