Ethical Implications of AI-Assisted Digital Phenotyping in Clinical Psychiatry

Abstract

The present paper explores how AI-assisted digital phenotyping (DP) reshapes the landscape of clinical psychiatry by reassigning authority over the collection and interpretation of mental health data and the interventions informed by such data. Digital phenotyping is an approach through which multimodal data streams and digital footprints of persons are collected and processed to support predictive and preventive care. While the assessment in the traditional setting explicitly involved two major stakeholders, i.e., specialists and patients, the integration of artificial intelligence (AI) in the field has altered this closed relationship. This, in turn, results in power asymmetry, where most of the power over the data is transferred to AI systems, and patients become mere objects of observation, and specialists lose their authority over data interpretation and clinical judgement. The paper discusses three major dimensions of authority that are challenged by this power asymmetry and proposes technical and ethical measures to safeguard legitimate human control and ownership over the collected data.