[Date]
Designing trust into AI recommendations
A model can be accurate and still fail in production if the person reviewing its output can't tell why it made a call. Some thinking on surfacing confidence and reasoning, drawn from working on RECs.
Topic staked out · draft pending