Harvard Researchers Unveil Adaptive AI Decision‑Support System

Harvard University scientists have created an adaptive artificial‑intelligence decision‑support system that tailors its assistance to individual users and contexts. The research addresses concerns that excessive reliance on generic AI tools can degrade human expertise and lead to poorer decisions. Using reinforcement‑learning techniques, the new system learns when and how much help to provide, improving joint human‑AI performance in online experiments with over a thousand participants. The study, led by PhD candidate Zana Buçinca and co‑authors Krzysztof Gajos and Maja Malaya, was published in ACM Transactions on Computer‑Human Interaction and will be presented at the ACM UIST conference later this year.