[2511.12832] From Passive to Persuasive: Steering Emotional Nuance in Human-AI Negotiation

Computer Science > Computation and Language arXiv:2511.12832 (cs) [Submitted on 16 Nov 2025 (v1), last revised 3 Mar 2026 (this version, v2)] Title:From Passive to Persuasive: Steering Emotional Nuance in Human-AI Negotiation Authors:Niranjan Chebrolu, Gerard Christopher Yeo, Kokil Jaidka View a PDF of the paper titled From Passive to Persuasive: Steering Emotional Nuance in Human-AI Negotiation, by Niranjan Chebrolu and 2 other authors View PDF Abstract:Large Language Models (LLMs) demonstrate increasing conversational fluency, yet instilling them with nuanced, human-like emotional expression remains a significant challenge. Current alignment techniques often address surface-level output or require extensive fine-tuning. This paper demonstrates that targeted activation engineering can steer LLaMA 3.1-8B to exhibit more human-like emotional nuances. We first employ attribution patching to identify causally influential components, to find a key intervention locus by observing activation patterns during diagnostic conversational tasks. We then derive emotional expression vectors from the difference in the activations generated by contrastive text pairs (positive vs. negative examples of target emotions). Applying these vectors to new conversational prompts significantly enhances emotional characteristics: steered responses show increased positive sentiment (e.g., joy, trust) and more frequent first-person pronoun usage, indicative of greater personal engagement. Our findings...