Microsoft Health Futures
Date: 13 February 2023
Time: 16:00 (CET)
Title: Human-Centred AI in Mental Health
Abstract: AI and machine learning (ML) promise significant transformations in the future delivery of mental healthcare, yet few works so far have moved beyond demonstrations of technical feasibility and algorithmic performance. To realize ambitious visions for how AI can contribute to clinical impact requires a closer design and study within their specific application context.
This talk draws on Project Talia, a recently retired, three-year research collaboration between Microsoft Research and SilverCloud Health – an evidence-based, digital, on-demand mental health platform that delivers Cognitive-Behavioral Therapy (iCBT) programs for patients with mild-to-moderate symptoms of depression, anxiety, and stress.
Focusing on the development of ML models to predict patient outcomes early within treatment, the talk specifically highlights the importance of human-centred research processes and the role of design and workflow integration for AI acceptance. It will discuss key choices that were made: (i) in defining what type of ML application to develop; and subsequently, (ii) in designing the front-end UI for communicating ML insights to healthcare professionals to assist specifically with risks of AI over-reliance. It closes with broader socio-technical considerations and challenges involved for the responsible design and workflow integration of ML within sensitive healthcare contexts.
Speaker Biography: Anja Thieme is a senior human-computer interaction researcher within Microsoft Health Futures (Cambridge, UK). Her research follows a human-centred, inter-disciplinary approach to the end-to-end development of innovative AI systems for use within real-world healthcare. More broadly, this work enables her to define strategy and processes that ensure AI technology is developed safely and responsibly. Amongst others, this is reflected in her contributions to Microsoft’s Aether working groups and the development of educational tools and other resources with members of the Office of Responsible AI (ORA) that seek to drive forward the adoption of responsible, human-centred AI practices.