Reducing Conversational Escalation in Large Language Model Dialogue with Nonviolent Communication Constraints
arXiv:2606.26106v1 Announce Type: cross Abstract: Large language models (LLMs) are increasingly used in emotionally charged situations involving interpersonal conflict, frustration, and distress. While prior safety research has focused on preventing explicit harms such as toxic or policy-violating content, less attention has been paid to conversational behaviors that may unintentionally escalate conflict. In this paper, we investigate whether LLMs can be guided toward more de-escalating dialogue behavior through lightweight prompt-level constraints derived from Nonviolent Communication (NVC).
As LLMs become more integrated into sensitive user interactions, the need for sophisticated emotional intelligence and conflict resolution capabilities is becoming acutely apparent, moving beyond basic safety filters.
This research addresses a critical gap in LLM safety and usability, enabling AI to navigate complex human emotions more effectively and reduce unintended negative outcomes in user interactions.
LLMs can now be theoretically guided to actively de-escalate emotional conflicts, shifting their role from mere information providers to potentially constructive conversational partners in fraught situations.
- · AI developers
- · Customer service industries
- · Mental health tech
- · Users interacting with AI
- · AI systems lacking sophisticated emotional intelligence
- · Platforms prone to AI-induced conflict escalation
LLMs trained with NVC principles could lead to more helpful and less frustrating AI interactions across various applications.
Public trust and acceptance of AI in sensitive roles could increase, accelerating AI adoption in critical sectors.
The application of NVC to AI could influence human-computer interaction design, fostering more empathetic and understanding digital communication norms.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.AI