AI 'Personality Engineering' lets researchers craft negotiators with warmth/dominance

A new preprint from MIT researchers Michelle Vaccaro and Jared Curhan introduces 'personality engineering,' a methodology that leverages AI agents to precisely control negotiator personality traits. The paper proposes using the interpersonal circumplex—a two-dimensional model mapping warmth and dominance—as the foundational coordinate system for negotiation experiments. This allows researchers to parameterize, manipulate, and evaluate personality factors with a precision, consistency, and scalability impossible with human subjects.

By removing human limitations (like emotional fatigue or social desirability bias), AI agents enable rigorous testing of canonical negotiation theories, such as the tension between empathy and assertiveness or between concern for self vs. others. Beyond academic research, the methodology provides a practical blueprint for building AI negotiation agents with tailored personalities—useful for simulations, training, or automated deal-making. The paper is available on arXiv (2605.20554) and could transform how we study and deploy AI in high-stakes conversations.