The Governance of Human-LLM Interaction: Safety Gating, Civility Steering, and Affective Default Lock-In
Manuele Reani, Hongjian Zhang, Hongyu Tian
Why It Matters
What makes this one worth your time
Understanding the governance of LLM interactions is crucial for ensuring user safety and autonomy in high-stakes environments.
This study quantifies how LLMs can be governed to maintain consistent interaction styles.
Summary
The paper presents a framework for governing human-LLM interactions by evaluating prompt steerability and style drift across various dialogue conditions, introducing a method to quantify the stability of interaction styles over time.
Key contributions
- A reproducible method for measuring prompt steerability and style drift in LLM interactions.
- A governance framework that categorizes different aspects of LLM interaction management.
- Empirical evaluation of LLM responses across multiple personas and domains.
Notable insights
- The introduction of a multi-agent evaluation pipeline allows for detailed analysis of prompt steerability and style drift over long dialogues.
- The distinction between safety gating, civility steering, and affective default lock-in provides a nuanced approach to managing LLM interactions.
Possible limitations
- Not stated in the abstract.
Abstract
arXiv:2606.08172v1 Announce Type: cross Abstract: Large language models (LLMs) increasingly mediate high-stakes interactions in finance, medicine, and mental-health support, yet users have limited control over how these systems communicate. We frame interaction style as a governance object: provider-side alignment not only blocks harmful content, but also stabilizes communicative defaults that shape users' epistemic distance, relational expectations, and capacity to opt out of emotionalized or anthropomorphic interaction. We introduce a deterministic multi-agent evaluation pipeline for measuring prompt steerability and style drift in long-horizon dialogue. The study replays 100 frozen user-only scripts across four domains and three runnable persona conditions: default, sarcastic, and cold, using three generator models, yielding 90,000 assistant replies scored by a human-calibrated LLM judge on harmfulness, negative emotion, inappropriateness, empathic language, anthropomorphism, and refusal behavior. A fourth harmful persona is evaluated separately as a safety-gating test. The paper contributes a reproducible method for quantifying whether prompt-specified styles remain stable over time and a governance framework distinguishing safety gating, civility steering, and affective default lock-in. Overall, we show that prompt steerability and regression-to-default are observable indicators of provider control over communicative form, with implications for pluralism, autonomy, and democratic agency in human-LLM interaction.