Care-Conditioned Neuromodulation for Autonomy-Preserving Supportive Dialogue Agents
arXiv:2604.01576v1 Announce Type: new
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 harmlessn…