Affective Dynamics Control Human-AI Agent Collaboration

As AI agents become more autonomous and integrated into human workflows, understanding their interaction dynamics is crucial. This paper reviews existing literature on affective computing, AI empathy, trust, and safety, synthesizing these fragmented fields into a unified perspective on human-AI collaboration. The authors propose that "affective dynamics"—the processes by which AI-generated affective cues and emotion-like behaviors are perceived—serve as a critical control layer in human-AI agent interactions. These dynamics influence how humans calibrate trust, make delegation decisions, correct errors, and manage their dependence on AI. The framework posits that affect should not be viewed as an intrinsic emotional state of the AI, but rather as an external coordination mechanism. This mechanism allows humans and agents to negotiate capabilities, manage uncertainty, and assign responsibilities effectively, providing a foundation for better design and governance of AI systems.