Get Started
Home
Topics
Search
Library
Research questionHow should AI assistants choose proposals when users cannot reliably evaluate them?A proposal may be useful for the task yet difficult for a user to judge, while a less immediately attractive proposal may reveal important preference information. Optimizing only for acceptance can therefore overlook how evaluability affects both assistance and preference learning.
AI
AI Agents
Alignment & Safety
Machine Learning
Latest papersRecent research connected to this question, newest first.Propose to Learn, Learn to Propose: Evaluability-Aware Assistance under Bounded RationalityThe source formalizes this as a hidden-parameter sequential assistance problem with a bounded-rational binary response model. Evidence comes from controlled graph simulations using a depth-2 Bayes-adaptive planner, with reported gains when evaluation cost is the bottleneck; broader deployment evidence is not provided.research paper · Sep 2, 2026
Related questions
How can AI agents adapt to each user’s evolving quality criteria on open-ended tasks?How should scientific agents be evaluated on underspecified, attachment-rich requests without ground truth?How can AI-assisted scoring reduce grading workload in national assessments without compromising human-judged scores or pass/fail decisions?How can we distinguish AI assistance from AI authorship when the prompting dialogue is unavailable?