Frictional Agent Alignment Framework: Slow Down and Don’t Break Things

ACL ARR 2025 February Submission4770 Authors

Abstract:

AI support of collaborative interactions entails mediating potential misalignment between interlocutor beliefs. Common preference alignment methods like DPO excel in static settings, but struggle in dynamic collaborative tasks where the explicit signals of interlocutor beliefs are sparse and skewed. We propose the Frictional Agent Alignment Framework (FAAF), to generate precise, context-aware "friction" that prompts for deliberation and re-examination of existing evidence. FAAF’s two-player objective decouples from data skew: a frictive-state policy identifies belief misalignments, while an intervention policy crafts collaborator-preferred responses. We derive an analytical solution to this objective, enabling training a single policy via a simple supervised loss. Experiments on three benchmarks show FAAF
outperforms competitors in producing concise, interpretable friction and in OOD generalization. By aligning LLMs to act as adaptive "thought partners"---not passive responders---FAAF advances scalable, dynamic human-AI collaboration.

Paper Type: Long

Research Area: Generation

Research Area Keywords: preference alignment in LLMs, collaborative dialogues, friction intervention

Contribution Types: Model analysis & interpretability, NLP engineering experiment, Theory

Languages Studied: English

Submission Number: 4770