Expert interfaces

What the research says

A nuance worth keeping: asking users to correct or label AI output can lower how accurate they judge the system to be, and lower their trust in it, even when their corrections measurably improve it. Task subjectivity moderates the effect. On objectively right-or-wrong tasks the hit shows up, on subjective ones it largely does not. So letting experts fix the output is not automatically a trust win, and the surface has to be worth the exposure it creates. 1Soliciting Human-in-the-Loop User Feedback for Interactive Machine Learning Reduces User Trust and Impressions of Model AccuracyDonald R. Honeycutt, Mahsan Nourani, Eric D. Ragan · HCOMP · 2020arxiv.org 2Human-in-the-Loop User Feedback Affects Perceived Accuracy and Trust, but Task Subjectivity MattersDonald R. Honeycutt, Mahsan Nourani, Eric D. Ragan · arXiv · 2026arxiv.org 3Guidelines for Human-AI InteractionSaleema Amershi, Dan Weld, Mihaela Vorvoreanu, Adam Fourney, Besmira Nushi, Penny Collisson, Jina Suh, Shamsi Iqbal, Paul N. Bennett, Kori Inkpen, Jaime Teevan, Ruth Kikin-Gil, Eric Horvitz · CHI · 2019doi.org