Cognitive Twins: Investigating Personalized Thinking Model Building and Its Performance Enhancement with Human-in-the-Loop

Researchers developed a Personalized Thinking Model (PTM) that creates 'cognitive twins' of learners by organizing educational data into a five-layer hierarchical structure using AI and machine learning. The system achieved 74-75% fidelity scores and positive user perception ratings, suggesting potential applications in AI-supported education systems.

The Personalized Thinking Model represents a significant advancement in educational AI by attempting to reverse-engineer and represent how individual learners think. Rather than treating education as one-size-fits-all, PTM uses large language models and clustering algorithms to build interpretable representations of student cognition across five abstraction levels—from concrete behavioral instances to abstract self-system values. This approach directly addresses a critical gap in AI-assisted learning: personalization at the cognitive level rather than merely content delivery.

The research builds on established educational psychology frameworks, specifically Marzano's taxonomy, which grounds the technical work in pedagogical theory. This combination of rigorous educational science with modern AI infrastructure creates a more trustworthy and interpretable system than black-box approaches. The seven-week study with 40 participants demonstrates real-world applicability, though sample size remains modest.

The results indicate meaningful technical success: F1 scores above 74% suggest the system captures meaningful patterns about learner thinking, while user perception ratings near 4.3/5.0 demonstrate that students recognize themselves in their cognitive twins. The semantic coherence pattern—increasing topic coherence while decreasing lexical overlap across abstraction layers—confirms the system performs genuine abstraction rather than mere summarization.

For the educational technology sector, PTM could enable adaptive learning systems that adjust to individual cognitive patterns rather than just learning pace. Organizations developing intelligent tutoring systems or personalized education platforms should monitor this research, as it offers a replicable methodology for building trustworthy cognitive models. The human-in-the-loop refinement process also suggests future commercial implementations could benefit from modest instructor oversight rather than requiring pure automation.

• →PTM achieves 74-75% fidelity in modeling individual learner cognition through hierarchical AI-driven analysis of educational journals.
• →The five-layer structure (behavioral instances to self-system values) enables interpretable, pedagogically-grounded cognitive representation.
• →Human-in-the-loop refinement improved user perception ratings marginally (4.26 to 4.30), suggesting the baseline system performs well without heavy human intervention.
• →Semantic coherence increases across abstraction layers while lexical overlap decreases, confirming genuine cognitive abstraction rather than simple summarization.
• →The approach could enable next-generation adaptive learning systems that personalize education to individual thinking patterns rather than just content difficulty.