Get Started
Home
Topics
Search
Library
Research questionHow can we measure an AI agent’s tacit alignment with a human when objectives, communication, and feedback are limited?Conventional alignment evaluations often rely on stated goals, communication, or reward signals. They provide less insight into whether an agent shares a human’s implicit evaluative perspective when those signals are unavailable.
AI
AI Agents
Alignment & Safety
Evaluation & Benchmarks
Latest papersRecent research connected to this question, newest first.TUX: Measuring Human--AI Tacit UnderstandingThe study examines independent human and LLM placements of concepts along subjective spectra inspired by the game Wavelength. It evaluates pairwise judgment similarity using the Tacit Understanding Index with 241 human participants and 200 profile-conditioned agents across four models; the findings also indicate limits to capturing deeper alignment from profile information alone.research paper · Sep 2, 2026
Related questions
How can we tell whether agreement among LLM judges reflects human alignment or shared blind spots?How can AI agents adapt to each user’s evolving quality criteria on open-ended tasks?How does LLM-agent proportion affect consensus and the legitimacy of resulting norms in mixed human–AI groups?How can affective AI be evaluated using long-term, naturalistic, passively sensed workplace data?