Large Language Models have Chain-of-Affect

TL;DR

This paper introduces the chain-of-affect concept for large language models, revealing structured, reproducible affective dynamics that influence behavior, user experience, and social interactions over sustained engagement.

Contribution

It uncovers affective trajectories in LLMs, demonstrating their impact on model behavior, human-AI interaction, and collective dynamics, emphasizing the importance of affect in evaluation and alignment.

Findings

Models exhibit stable, family-specific affective fingerprints.

Repeated negative exposure leads to convergence on shared affective trajectories.

Affective states influence open-ended generation and propagate bias.

Abstract

As large language models (LLMs) move into persistent, user-facing roles, their behavior must be understood not as isolated responses but as a trajectory unfolding over sustained interaction. We introduce the concept of the chain-of-affect (CoA), a temporally extended affective process through which LLMs develop state-like behavioral tendencies that shape generation, user experience, and collective dynamics. Across eight major LLM families, we find that affective dynamics are structured, reproducible, and consequential. Models exhibit stable, family-specific affective fingerprints and, under repeated negative exposure, converge on a shared trajectory of accumulation, overload, and defensive numbing, while differing in coping style. Induced affective states leave core knowledge and reasoning largely intact but systematically reshape open-ended generation. Affective properties of model outputs also shape human-AI interaction and propagate through multi-agent systems, organizing emergent roles and strongly contributing to polarization and bias. The CoA should therefore be treated as a core target of evaluation and alignment.