Skip to content

Partially specified structural causal model (PSCM) protocol for educational assessment

Research Method · All research methods
Evidence · 1 claim (1 mixed) · 1 study (1 design), q2 · 0 of 1 report an effect size · 1 claim rests on one study

Description

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. As the article states, "endogenous variables — students' answers and hint usage — are the observed variables of our model, generated by the latent exogenous variables skill, luck, and help-seeking". Structural equations are elicited from domain experts in purely logical form, and the resulting graph is acyclic by construction, making the model semi-Markovian and ensuring more tractable inference. The framework explicitly supports interventional and counterfactual reasoning beyond associative probabilistic learner models.

Accounts

  • 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. As the article states, "endogenous variables — students' answers and hint usage — are the observed variables of our model, generated by the latent exogenous variables skill, luck, and help-seeking". Structural equations are elicited from domain experts in purely logical form, and the resulting graph is acyclic by construction, making the model semi-Markovian and ensuring more tractable inference. The framework explicitly supports interventional and counterfactual reasoning beyond associative probabilistic learner models. (Francesca Mangili et al. (2026))
  • Hypothetical data-generating model with trait and state components for diagnosing bias in instruction-effect estimation: The article proposes representing, via graphs, a hypothetical data-generating model containing all observed and unobserved variables affecting the outcome at T1 and T2 and the instruction variable. As the article states, "we use graphs to represent the state component, the trait component, their effects on the variables of interest (outcome at T1, outcome at T2, and instruction), and the instruction effect, denoted by δ". Researchers formulate assumptions about which components influence which variables, then deduce for which scenario the true effect δ is mirrored by the ANCOVA or CHANGE coefficient. Bias is defined as the difference between the true effect in the data generation and the estimated effects. (Köhler et al. (2021))

Claims

-

Key Sources

  • Francesca Mangili, Alessandro Antonucci, Rafael Cabañas de Paz. (2026). Causal Modelling of Support Interventions for Student Competency Assessment. arXiv. https://arxiv.org/abs/2608.24632
  • Köhler, C., Hartig, J., & Naumann, A. (2021). Detecting Instruction Effects—Deciding Between Covariance Analytical and Change-Score Approach. Educational Psychology Review. https://doi.org/10.1007/s10648-020-09590-6
Page metadata
Field Value
type research-method
id partially-specified-structural-causal-model-pscm-protocol-for-educational-assessment
title Partially specified structural causal model (PSCM) protocol for educational assessment
description 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.
status review
generated by process:settle-candidates on 2026-10-10
trust unverified — no one has confirmed this page's content

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