AI Cannot Feel the Trap: Narcissism, Manipulation and the Future of Negotiation Technology

Abstract

As AI increasingly assists negotiation processes, an overlooked vulnerability emerges: reliance on cooperative and rational-choice assumptions leaves such systems ill-equipped to recognise and manage narcissistic and high-conflict personalities. Both traditional and AI-driven models anticipate stable interests, good-faith engagement and mutual incentives to reach agreement. Current AI negotiation tools rely on trust-building cues, sentiment analysis and affect recognition to interpret interpersonal dynamics. Yet narcissistic and high-conflict personalities may weaponize negotiation processes to preserve a grandiose self-image, punish counterparts and seek psychological dominance rather than substantive agreement. Their tactics – charm, provocation, reputation attacks, procedural manipulation and emotional destabilisation – exploit and distort negotiation dynamics. In these contexts, AI negotiation systems misread hostility as information, conflate manipulation with cooperation and risk inadvertently supporting abusive strategies. This paper bridges psychological research on narcissism in conflict with scholarship on AI-assisted negotiation and AI ethics. It identifies structural vulnerabilities in negotiation technology when confronted with adversarial personality dynamics and proposes design principles including behavioural-risk indicators, safeguards against manipulative tactics and ethical boundaries in AI-supported bargaining. As AI-driven negotiation expands across legal, commercial, diplomatic and humanitarian domains, addressing these psychological asymmetries is essential for fairness, safety and system integrity.