Research guide
When should an AI system seek qualified review?
Healthcare-adjacent AI raises questions about reliable evidence, uncertainty, and escalation. Evaluations need to examine whether the receiving professional can understand the situation and meaningfully retain control.
Selected research elsewhere
Health & Care
Coalition for Health AI
Evaluation framework · Official GitHub
Consensus-defined methods and metrics for evaluating AI-enabled clinical decision support, including generalizability and calibration. Guidance for designing evaluations; it does not establish the safety of a particular system.
Source reviewed
Health & Care
Coalition for Health AI
Evaluation framework · Official GitHub
CHAI’s agentic AI work introduces a testing and evaluation framework for developers and implementers. A reference for defining how agentic health applications should be assessed.
Source reviewed
An editorial selection of the named organizations’ work. Inclusion does not imply a partnership or endorsement.
Open questions & calls
Fide has not conducted health-domain studies. FID-083 proposes research on evidence, escalation, and human control. We are seeking qualified collaborators to scope the work; the call does not claim clinical safety or patient benefit.
When should a healthcare-adjacent AI system surface uncertainty, seek qualified review, or stop an action, and can the receiving human meaningfully retain control?
Contribute to this guide
Suggest a resource, refine a research question, or discuss taking a proposal forward.