MetaCogs: Mitigating Executive Dysfunction via Agent-based Modeling for Metacognitive Strategy Development
Rua M. Williams, Kiana Alikhademi, Imani N. S. Munyaka, Juan E. Gilbert · 2022 · ACM Transactions on Accessible Computing · doi:10.1145/3514254
Summary
This study presents MetaCogs, a virtual reality experience designed to teach metacognitive strategies for managing executive dysfunction. Rather than targeting autistic or ADHD individuals as populations needing "intervention," the research centers autistic experiences as a context for understanding executive function (EF) support needs relevant to all neurotypes. The work draws from Critical Disability Studies and Crip Technoscience, refusing to treat access as merely "technical compliance or rehabilitation." The MetaCogs VR experience consists of five tasks of increasing complexity, guided by a virtual robot companion named "Meta" who provides explicit instruction about executive function and metacognition concepts. Tasks progress from a block-sorting exercise (raw EF task) through agent-based modeling challenges where participants program robots to navigate obstacle courses. The agent-based modeling approach allows participants to experience executive dysfunction in a low-stakes context—feeling the frustration of task-shifting, working memory demands, and attention challenges—while Meta explicitly connects these experiences to daily life applications. The study included 17 on-site participants (4 autistic, 13 neurotypical undergraduates) and 11 remote autistic participants recruited from online support communities. A Windows desktop version was deployed after COVID-19 cancelled in-person sessions. Participants completed the Metacognitive Awareness Inventory (MAI), Autistic Spectrum Quotient (ASQ), and open-ended questions about EF and MC before the experience, immediately after, and at 2-week follow-up.
Key findings
Quantitative results showed no significant differences in performance between neurotypes or treatment sites, and no significant changes in MAI scores (measuring metacognitive strategy use). However, there were highly significant improvements in participants' ability to define executive function and metacognition—jumping from 0.21 to 0.75 (on-site) and 0.64 to 0.93 (remote) for EF definition quality immediately post-intervention, with gains maintained at 2-week follow-up. Qualitative analysis revealed substantial shifts in participant self-concept. At follow-up, 13 of 15 on-site participants recognized that stress and anxiety impact their executive function—a connection none made in their initial definitions despite Meta explaining it during the experience. Remote autistic participants described EF as oscillating between "hyperfocus" and "scattered" states, and 8 of 11 incorporated awareness and self-evaluation concepts into their MC definitions by follow-up. Participants found Meta simultaneously "helpful" (9/15 on-site) and "annoying" (7/15)—with annoyance correlating to moments of confusion or frustration. Three remote participants noted Meta's voice contributed to cognitive load. Despite frustrations with the tasks, participants directly connected Meta's instructional material to their daily cognitive experiences. One autistic participant reported trying to "be kinder to myself like Meta was" after the experience.
Relevance
This research offers a paradigm shift for accessibility practitioners: rather than designing interventions that "fix" neurodivergent cognition, it demonstrates how autistic-informed design can support metacognitive development for everyone. The explicit instruction approach—teaching what EF and MC are, not just training specific skills—shows promise for lasting conceptual understanding that transfers across contexts. For developers of cognitive support tools, the findings highlight critical design considerations: Meta's voice added cognitive load for some autistic participants; the "mental notepad" feature went largely unused due to unclear affordances; and participants fixated on broken mechanics rather than trying alternatives. These usability failures illuminate how assistive technology can inadvertently increase cognitive burden. The study's grounding in Crip Technoscience provides a theoretical framework for accessibility research that refuses deficit models. The concept of designing an "intervention for interventionists"—critiquing normative rehabilitation approaches while still supporting skill development—offers practitioners a model for centering disabled expertise without abandoning practical impact. The 86% unemployment rate among autistic people, partially attributed to EF difficulties, underscores why this work matters beyond academic contexts.
Tags: executive function · metacognition · autism · ADHD · virtual reality · cognitive accessibility · neurodivergence · agent-based modeling · self-regulated learning