Authors: (including presenting author): :
JSM Leung(1), CP Cheung(1), NW Sezto(1), N Hui(1), HH Wong(2)
Affiliation: :
(1)Occupational Therapy Department, United Christian Hospital
(2)Department of Psychiatry, United Christian Hospital
Keyword 1: :
Functional Cognition
Keyword 2: :
Generative AI
Keyword 3: :
Cognitive impairment
Keyword 4: :
Mental Illness
Introduction: :
Cognitive impairment is a major barrier to daily functioning in mental illness. Up to 85% of people with schizophrenia experience significant cognitive deficits that interfere with daily living (Ji et al., 2025). And around 40% of individuals with mood disorder (Decker et al., 2017; Zhao et al., 2022) reported difficulties in managing instrumental activities of daily living (IADL). Traditional IADL training remains the cornerstone of evidence-based interventions for clients with mental illness which provide structured skill-building and rely on the therapeutic alliance to foster long-term cognitive and behavioral change. However, generative AI can complement these methods by delivering real-time, personalized support in everyday contexts.
Objectives: :
To assess the feasibility and impact of GenAI in facilitating community re-integration for adults with mental illness.
Methodology: :
This study involved a transdiagnostic convenience sample (n=10; 9 male, 1 female) from Yung Fung Shee Psychiatric Day Hospital, from April to June, 2025. Diagnoses of participants included schizophrenia, bipolar disorder, psychotic depression and mild intellectual disability with mood disorder, etc. Participants completed 13-sessions IADL training with facilitation of Grok (a GenAI model), including health management, budgeting, time management, route planning and meal preparation, which involved using AI-generated recipes for cooking healthy meals and demonstrating skills to peers. Outcomes were measured by the Technology Acceptance Model (TAM) questionnaire for “perceived usefulness” and “ease of using” of GenAI, and Adapted Self-Efficacy for Social Participation Scale (SOSA) for self-efficacy in community re-integration. Qualitative data were gathered from patients’ feedback.
Result & Outcome: :
The adapted SOSA scale showed a preliminary Cronbach’s Alpha of 0.85. Subsequent paired-samples t-tests showed significant improvements in both TAM (p < .01) and Adapted SOSA in instrumental (p < .01), managerial (p < .01) and cultural (p < .01) domains. This indicated that participants perceived GenAI as both useful and easy to use while significantly boosted their self-efficacy in community re-integration. Qualitatively, participants viewed GenAI as an effective cognitive aid that helped lower cognitive demand for tasks like route planning and budgeting, and transformed tasks into manageable activities. However, sustainability of using this application was limited by technical issues, e.g. input modalities and cultural adaptation in generated content, underscoring the importance of human support. This study provided preliminary evidence of GenAI's feasibility and efficacy as a cognitive adjunct for individuals with mental illness by showing promising results in perceived functional capacity and community re-integration. This study further demonstrated that combining AI planning with hands-on execution promoted skill mastery rather than dependency on digital tools. However, technical and cultural barriers highlighted the need for human oversight. These findings suggested that occupational therapists play a vital role in enhancing clients’ competence in understanding, use and evaluating GenAI technologies and composing precise commands for GenAI as contemporary community-living skills. GenAI may serve as a valuable adjunct to enhance practice opportunities and improve accessibility. Future research should prioritize rigorous trials and robust safety frameworks to ensure its effective integration with traditional training.
