Artificial intelligence for neurodiversity

Beyond the extreme risks of artificial intelligence (AI)1 discussed by Sorich and colleagues,2 algorithmic bias presents unique challenges to patients with mental disorders, particularly neurodevelopmental disorders such as autism spectrum disorder and attention deficit/hyperactivity disorder (ADHD). This bias arises because large language models are trained on “big data” from “average” patients, focusing on single diseases without incorporating the diversity of comorbid patient subgroups.We urge AI innovators and clinical users to view cognitive diversity as an opportunity for precision medicine rather than as noise in machine learning models. AI could:Target comorbid patients. Although AI models for cancer and cardiovascular diseases are well developed,3 “big data” need to be “small” enough to customise for neurodivergent patients, while preserving privacyIntegrate physical and mental health information. Electronic medical record systems could nudge specialists to consider conditions like autism, conduct disorder, or post-traumatic stress disorder in doctor-patient interactionsExpedite the recognition of depression and anxiety and…
Read Original Article: Artificial intelligence for neurodiversity »