Care-Conditioned Neuromodulation for Autonomy-Preserving Supportive Dialogue Agents

2026-04-01

AI Summary

This work addresses the lack of explicit modeling of user autonomy violations—such as dependency and overprotection—in supportive conversations by current large language models. The authors propose a state-dependent control framework that, for the first time, integrates neuromodulatory mechanisms into dialogue alignment. By learning scalar signals from user states and contextual cues, the framework dynamically modulates response generation. It incorporates a structured representation of user states to construct an autonomy-preserving utility function and introduces a new benchmark for relational failure modes in multi-turn supportive dialogues. Experiments on this benchmark demonstrate that the proposed method improves autonomy-preserving utility by 0.25 and 0.07 over supervised fine-tuning and preference optimization baselines, respectively, while maintaining comparable levels of supportiveness. Both human evaluations and automatic metrics corroborate these findings.

Abstract

Large language models deployed in supportive or advisory roles must balance helpfulness with preservation of user autonomy, yet standard alignment methods primarily optimize for helpfulness and harmlessness without explicitly modeling relational risks such as dependency reinforcement, overprotection, or coercive guidance. We introduce Care-Conditioned Neuromodulation (CCN), a state-dependent control framework in which a learned scalar signal derived from structured user state and dialogue context conditions response generation and candidate selection. We formalize this setting as an autonomy-preserving alignment problem and define a utility function that rewards autonomy support and helpfulness while penalizing dependency and coercion. We also construct a benchmark of relational failure modes in multi-turn dialogue, including reassurance dependence, manipulative care, overprotection, and boundary inconsistency. On this benchmark, care-conditioned candidate generation combined with utility-based reranking improves autonomy-preserving utility by +0.25 over supervised fine-tuning and +0.07 over preference optimization baselines while maintaining comparable supportiveness. Pilot human evaluation and zero-shot transfer to real emotional-support conversations show directional agreement with automated metrics. These results suggest that state-dependent control combined with utility-based selection is a practical approach to multi-objective alignment in autonomy-sensitive dialogue.

Problem

• autonomy preservation
• supportive dialogue
• relational risks
• dependency reinforcement
• coercive guidance

Innovation

• Care-Conditioned Neuromodulation
• autonomy-preserving alignment
• relational failure modes
• utility-based reranking
• state-dependent control