Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
The paper introduces Care-Conditioned Neuromodulation (CCN), a framework for training dialogue agents to balance helpfulness with user autonomy. CCN uses a learned scalar signal derived from user state and dialogue context to condition response generation and selection, optimizing a utility function that rewards autonomy support while penalizing dependency and coercion. Experiments on a new benchmark of relational failure modes show that CCN improves autonomy-preserving utility compared to supervised fine-tuning and preference optimization, with promising results in pilot human evaluations and zero-shot transfer.
LLMs can learn to balance helpfulness with user autonomy by conditioning response generation on a learned "care" signal, mitigating relational risks like dependency and coercion.
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.