Modeling metacognition and executive functions in the metacognitive wisconsin card sorting test using the neuropsychological digital-twin method.

A new computational model integrates executive functions and metacognition to explain goal-directed behavior. The model, tested with the Metacognitive Wisconsin Card Sorting Test, reproduces data from healthy individuals and those with Anorexia Nervosa or Schizophrenia. Results reveal hidden similarities between the disorders—motivational impairment and over-confidence—alongside differences: perseveration and poor self-improvement in Anorexia Nervosa, distraction and poor self-evaluation in Schizophrenia. Simulations predict that metacognitive-based psychotherapy benefits differ by condition, highlighting the need for personalized interventions. The digital-twin method offers a framework for neuropsychological profiling and prediction.

Study at a glance

Abstract

Executive functions rely on goal-directed manipulation of representations, while metacognition reflects the evaluation and control of one’s own representations. Several studies have examined these processes separately, but none formalise the neuro-representational computations underlying their interaction during goal-directed behaviour. This gap prevents comprehensive frameworks and extended model-based neuropsychological investigations. Here we address these issues by introducing a neuropsychological digital-twin method - a translational modelling framework that integrates clinical and experimental data, theoretical formalisation, and computational modelling for neuropsychological profiling and prediction. We formalised the three-component theory of metacognitive and flexible goal-directed cognition - grounded in theoretical and neuroscientific literature - and developed a neuro-inspired computational model tested with a standard neuropsychological task (Metacognitive Wisconsin Card Sorting Test, Meta-WCST). We further corroborated the proposal by reproducing experimental data from healthy controls and psychiatric populations (Anorexia Nervosa and Schizophrenia). Finally, we generated three digital-twins - computational models fitted to human data and reproducing behavioural and neuro-cognitive features - for neuropsychological profiling and intervention prediction. Our results support an integrated framework of executive functions and metacognition in healthy and pathological goal-directed behaviour and provide the first theory-based computational model of the Meta-WCST. They also reveal that Anorexia Nervosa and Schizophrenia share hidden cognitive and metacognitive similarities (motivational impairment and over-confidence) alongside differences (perseveration and poor self-improvement in the former; distraction and poor self-evaluation in the latter). Consistently, simulations predict differential benefits from metacognitive-based psychotherapy, highlighting the importance of personalised interventions. Finally, our contributions have implications for cognitive science (e.g., consciousness studies) and emerging technologies (digital-twin healthcare and autonomous robotics).