Generative AI collective behavior needs an interactionist paradigm

TL;DR

The paper tackles understanding the collective behavior of Gen-AI agents initialized with large pre-trained priors and capable of in-context adaptation, framing the problem as a social–cultural dynamic among interacting models. It proposes an interactionist paradigm built on four pillars—an interactionist theory, causal inference, information theory, and a sociology of machines—to analyze how internal priors and social interactions jointly shape emergent group behavior. It contrasts Gen-AI collectives with traditional MARL, arguing that in-context learning, rather than weight updates, drives adaptation and second-order collective emergence, and it provides a formal model and benchmarking directions to study these phenomena. The authors outline concrete research directions—interactionist benchmarks, causal identification designs, information-theoretic analyses, and empirical sociology—to guide design, governance, and safe deployment of Gen-AI collectives in real-world settings.

Abstract

In this article, we argue that understanding the collective behavior of agents based on large language models (LLMs) is an essential area of inquiry, with important implications in terms of risks and benefits, impacting us as a society at many levels. We claim that the distinctive nature of LLMs--namely, their initialization with extensive pre-trained knowledge and implicit social priors, together with their capability of adaptation through in-context learning--motivates the need for an interactionist paradigm consisting of alternative theoretical foundations, methodologies, and analytical tools, in order to systematically examine how prior knowledge and embedded values interact with social context to shape emergent phenomena in multi-agent generative AI systems. We propose and discuss four directions that we consider crucial for the development and deployment of LLM-based collectives, focusing on theory, methods, and trans-disciplinary dialogue.

Figures (3)

• Figure 1: Conceptual description of the shift from simple Gen-AI agents to a collective of Gen-AI agents. (1) On the top left, we visualize an initial phase in which an agent learns via pre-training, supervised fine-tuning (SFT), and reinforcement learning from human feedback (RLHF). Then, two different in-context experiences are represented: (2) on the bottom left, the agent is employed in in-context learning (ICL) to solve a given task, which leads to individual behaviors; (3) on the right, instead, four agents trained according to (1) are involved in interactive ICL tasks. The interaction between the Gen-AI agents can lead to emergent complex collective behaviors that are the byproduct of individual and situational conditions.
• Figure 2: Overview of the LLM development pipeline. The model undergoes four sequential learning phases: (1) pre-training on web-scale generalist data via next-token prediction to acquire parametric knowledge; (2) supervised fine-tuning (SFT) on curated in-domain data to learn task-specific skills such as reasoning and instruction following; (3) alignment through reinforcement learning from human feedback (RLHF), using preference data to optimize for helpfulness, faithfulness, and truthfulness; and (4) an interactive deployment phase, where agents exhibit adaptive behavior through interactive in-context learning (ICL).
• Figure 3: Four areas in which our paradigm would be critical: (a) the study of emergent phenomena, exemplified by the analysis of shared meaning of concepts in debates, through the interactionist lens; (b) the evaluation policies applied to MAS via causal inference; (c) the identification of sources of behavioral diffusion in MAS via information-theoretic insights and causal inference design; (d) the development of theories via empirical scrutiny building on sociological frameworks originally designed for humans.