AI as Cognitive Models in Education: Theory, Synthesis, and Mechanistic Insight

This Topical Collection, AI as Cognitive Models in Education: Theory, Synthesis, and Mechanistic Insight, is explicitly oriented toward advancing theoretical and integrative work in educational psychology. Our primary audience is the educational psychology community. We invite scholarship that uses AI-based models (e.g., artificial neural networks, agentic AI) as vehicles for theory development, testing, and refinement—anchoring computational innovations in established constructs such as memory, attention, problem-solving, motivation, and metacognition. Rather than reporting black-box performance gains or surveying the broader AI-in-education landscape, submissions should provide interpretable, mechanistic accounts of how students think and learn and synthesize insights across cognitive frameworks.

AI technologies (from neural networks and reinforcement learning agents to large language models) offer unprecedented opportunities to test precise theoretical ideas about the mind. Instead of remaining at an abstract level, researchers can now instantiate cognitive theories in AI models and examine how learning unfolds step by step. For example, an AI model trained to solve problems or learn concepts can be analyzed to reveal the specific mechanisms by which it acquires knowledge or fails to do so, potentially mirroring human strategies and errors. Such approaches can make our theories more sophisticated and enable new syntheses between AI and cognitive science. However, to truly benefit educational psychology, these AI models must be interpretable and aligned with psychological constructs. Therefore, this collection emphasizes transparency, interpretability, and theoretical fidelity: AI-based models should not only predict learning outcomes but also explain why and how those outcomes occur in terms that resonate with educational theory. Submissions should stress the conceptual insights gained from AI models—how they inform cognitive load, strategy use, knowledge construction, or other learning-related processes—rather than just reporting performance gains. Ultimately, this Topical Collection will showcase how AI, used as cognitive models, can advance theory development, integrate fragmented research findings, and guide practical innovation in educational psychology.

Topics of Interest:

We welcome submissions addressing a broad range of topics within this theme. Emphasis is on theoretical, conceptual, and integrative contributions, but we will also consider selected empirical or simulation studies that yield clear theoretical insights. Topics of interest include, but are not limited to:

· Integrative Theoretical Frameworks: Proposals of new frameworks or models that bridge AI techniques (e.g. deep neural networks, cognitive architectures, large language models) with established educational psychology theories. For instance, papers might explore how AI learning algorithms can instantiate principles from working memory theory, cognitive load theory, self-regulated learning, or other paradigms, thereby linking computational processes with human cognitive constructs.

· Reviews and Syntheses: Comprehensive review articles or meta-analyses synthesizing the literature on AI as models of human learning and cognition. This could involve critical analyses of past research on intelligent tutoring systems, student modeling, or cognitive simulations in education, identifying what technical advances mean for educational psychology and highlighting gaps where theory and AI findings diverge.

· Mechanistic Insights from AI Models: Studies that use AI systems to reveal how learning happens at a mechanistic level. This might include analyses of trained AI models to uncover strategies or representations analogous to those of human learners, shedding light on processes like problem-solving steps, misconception formation, memory retention, or transfer of learning. Such contributions should demonstrate how examining an AI’s internal workings (e.g. weight patterns, decision rules, generated examples) provides explanatory power for human learning phenomena.

· Alignment with Cognitive Constructs: Investigations into integrating specific psychological constructs into AI-driven learning models. For example, approaches to incorporate cognitive load, attention, or metacognitive monitoring into machine learning models of student performance, or AI simulations that factor in motivational states and affect. Topics may include how to design AI agents whose behavior can be mapped onto elements of cognition (e.g. an AI tutor that simulates a student’s forgetting curve or self-regulation process), and what this reveals about the construct’s role in learning.

· Interpretability and Explainable AI in Educational Contexts: Methods and case studies for making AI models in education transparent and interpretable to researchers and educators. We invite work on explainable AI techniques that connect model behavior to human-understandable reasoning (for example, breaking down a neural network’s function in pedagogical terms). Contributions may discuss how interpretable AI models can support or refine psychological theories, and frameworks for ensuring that AI’s decisions or recommendations can be explained via educational psychology principles.

· Empirical Illustrations of AI as Cognitive Models: Small-scale empirical studies or simulation experiments that demonstrate the use of AI to test hypotheses about learning. This could involve, for instance, comparing the learning curve or error patterns of an AI model to those of students on the same task to validate a cognitive theory, or human-subject experiments where AI-suggested mechanisms are evaluated. While full-scale interventions are outside the primary scope, we will consider proof-of-concept studies that empirically illustrate how AI-driven models can emulate or illuminate human learning processes.

· Critical and Ethical Analyses: Conceptual papers examining the assumptions, limitations, and ethical implications of using AI as models of human cognition in education. This includes discussions of where AI models align or misalign with human cognitive architecture, the risks of misinterpreting AI behavior as human behavior, and the importance of addressing bias and fairness when AI frameworks are used to represent student thinking. Such contributions should still maintain a focus on theoretical and mechanistic understanding (as opposed to broad ethical issues), for example by critiquing the theoretical validity of certain AI models as proxies for learners or the conditions under which AI insights can legitimately inform educational theory.

Please note: This collection does not consider empirical studies that merely evaluate human use of LLMs (e.g., comparing outcomes when students or teachers use ChatGPT). Submissions must go beyond usage evaluation to offer clear theoretical or mechanistic insights—for instance, by using AI systems as models of cognition, learning, or behavior, or by interpreting their internal processes in psychologically meaningful ways.

Expected Contribution Types:

Authors may submit various types of manuscripts, with a primary focus on rigorous theoretical and integrative work: Theoretical and Conceptual Papers, Review and Synthesis Articles, Methodological or Perspective Pieces, Model-based Empirical, Intervention, and Simulation Studies (Theory-driven).

Submission & Requirements

• Where to submit: Through Educational Psychology Review’s submission system, via the journal homepage. Select the Topical Collection “AI as Cognitive Models in Education” at submission. Pre-submission enquiry is possible at wang@essb.eur.nl.
• Format: Follow EDPR author guidelines.
• Length: Around 12,000 words, consistent with journal standards.
• Theoretical alignment: Papers must be clearly grounded in educational psychology theory, with AI used to explain or test learning-related constructs—not just to showcase technology.
• Mechanistic insight & interpretability: Emphasize how the AI model explains learning processes (e.g., mechanisms, strategies, representations) in psychologically meaningful terms and describe any interpretability methods used.
• Practical implications: Include implications for instruction, assessment, adaptive systems, or policy, even in primarily theoretical work.
• Ethics & quality: Adhere to ethical standards (e.g., IRB approval where relevant, transparent methods). Submissions are expected to meet high scholarly and methodological standards.

Review Process

All manuscripts undergo EPR’s standard double-blind peer review:

• Initial screening by the guest editors (and, where appropriate, the Editor-in-Chief) for fit and basic quality.
• External review by at least two expert reviewers (typically three), with conflicts of interest avoided.
• Guest editors synthesize reviews and recommend decisions; final decisions are made in consultation with the journal’s editorial leadership.
• Revised manuscripts typically undergo one or more additional rounds of review, as needed to ensure quality and alignment with the collection's standards.
• Manuscripts authored or co-authored by guest editors are handled by an independent handling editor without their involvement.