Align When They Want, Complement When They Need! Human-Centered Ensembles for Adaptive Human-AI Collaboration
Hasan Amin, Ming Yin, Rajiv Khanna · 2026
Read Align When They Want, Complement When They Need! Human-Centered Ensembles for Adaptive Human-AI Collaboration by Hasan Amin, Ming Yin, Rajiv Khanna (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-20104.txt.
Description
An AI assistant faces a dilemma. Matching a human's approach builds trust but can reinforce mistakes. Complementing human expertise boosts performance but can cost trust exactly when advice matters most. A 2026 arXiv paper by Hasan Amin and two co-authors names that tension as an inherent limit of training one AI model to assist human decisions.
To overcome it, the authors introduce a human-centred adaptive ensemble. It toggles between two specialist models, one aligned and one complementary, based on context cues. The routing mechanism, called the Rational Routing Shortcut, is simple yet provably near-optimal. Theory explains why the ensemble works and when it yields the largest benefits. Experiments on simulated and real-world data show people assisted by the adaptive ensemble achieve much higher performance. That edge holds against single models trained either for solo performance or for human-AI team performance.
Catalog record
| CREATOR | Hasan Amin, Ming Yin, Rajiv Khanna |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 87 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-20104