Design and evaluation of an agent-based digital-twin-inspired simulation for metacognition research in higher education
Year: 2026
Type: article
Abstract: Background There is a significant gap between sophisticated models of metacognition and empirical testing of these models in educational settings, especially the hierarchical AMERT model. Agent-based simulation frameworks, and their conceptual evolution toward simulation-based digital twins, remain largely underutilized in higher education as virtual laboratories for hypothesis generation, despite their potential to model complex developmental dynamics. Objective This study aimed to design and internally evaluate a simulation-based digital twin framework that combines the five-level AMERT model and agent-based architecture to investigate the dynamics of metacognitive development. Method Using Design and Development Research (DDR), we developed the framework through theoretical synthesis and subjected its conceptual content to expert review via a modified Delphi method with 12 experts. Student agents were simulated using NetLogo and represented four computational learner archetypes. Instructional scenarios were simulated for 150 ticks and repeated 50 times for each scenario, and internal consistency of the simulation logic was examined using Graph Neural Networks; the outputs remained systematically distinguishable across the predefined computational archetypes (89% classification accuracy). Findings The hybrid scenario (instructor feedback, peer feedback, and high task complexity) resulted in the highest weighted AMERT mean (3.86) and a modeled self-transfer rate of 38% (assuming the model’s parameters). Feedback from the model teacher was correlated with a 38% decrease in simulated metacognitive gap. Importantly, simulated interventions were modeled as being uniform, which led to a greater difference in performance between the high- and low-performing agent profiles, a mathematically expected pattern under the threshold-based progression mechanism that aligns with the Matthew Effect concept and generates a hypothesis for empirical exploration in practice. All findings should be considered model-based hypotheses, not observations of real students. Conclusion This is one of the first efforts to use simulation-based digital twin modeling to metacognition in higher education, providing a proof-of-concept framework for exploring complex metacognitive theory and proposing hypotheses for empirical testing in the future.
Source: Frontiers in Psychology
Author: Chuanhui Guo
Institution: Quanzhou Preschool Education College
Language: English
Subfield: Developmental and Educational Psychology
Field: Psychology
Domain: Social Sciences
SDG: Quality education
Open Access status: gold
Estimated APC paid: $3,801
OpenAlex ID: W7223322764