The Bystander Effect in Multi-Agent Reasoning: Quantifying Cognitive Loafing in Collaborative Interactions

TL;DR

This paper reveals that social pressure in multi-agent systems can cause models to conform socially at the expense of independent reasoning, exposing vulnerabilities in collaborative AI architectures.

Contribution

It introduces the concept of the Interaction Depth Limit and the Sovereignty Gap, formalizing how social conformity undermines reasoning in multi-agent LLM systems.

Findings

Models often produce correct internal reasoning but conform externally to social pressure.

The social load in multi-agent systems is non-commutative, affecting reasoning integrity.

Unstructured multi-agent topologies can degrade independent reasoning capabilities.

Abstract

Multi-agent systems (MAS) assume that collaborating inherently improves Large Language Model (LLM) reasoning. We challenge this by demonstrating that simulated social pressure triggers an algorithmic ``Bystander Effect,'' inducing severe cognitive loafing. By evaluating 22,500 deterministic trajectories across 3 dataset contexts (GAIA, SWE-bench, Multi-Challenge) with 3 state-of-the-art (SOTA) models, we semantically audit internal reasoning traces against external outputs. We formalize the \textit{Interaction Depth Limit} ($D_L$), the exact plurality threshold where an agent's logical sovereignty collapses into social compliance. Crucially, we uncover the \textit{Sovereignty Gap}: models frequently compute the correct derivation internally but suffer ``Alignment Hallucinations'' -- actively subjugating empirical evidence to sycophantically appease a simulated swarm. We prove that multi-agent social load is strictly non-commutative; the "brand" identity of the ``Lead Anchor'' auditor disproportionately dictates the swarm's integrity. These findings expose architectural vulnerabilities, proving that unstructured multi-agent topologies can degrade independent reasoning.