Generative AI collective behavior needs an interactionist paradigm

TL;DR

This paper emphasizes the importance of adopting an interactionist paradigm to understand the collective behavior of large language model agents, considering their social priors, adaptation, and emergent phenomena, with implications for societal risks and benefits.

Contribution

It introduces the need for a new interactionist framework and outlines four key directions for advancing theory, methods, and interdisciplinary dialogue in multi-agent generative AI systems.

Findings

Highlights the role of pre-trained knowledge and social priors in LLM behavior

Proposes an interactionist paradigm for analyzing multi-agent AI systems

Suggests four crucial directions for future research and deployment

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.