Adaptive Accountability in Networked MAS: Tracing and Mitigating Emergent Norms at Scale

TL;DR

This paper introduces the Adaptive Accountability Framework (AAF), a runtime system for large-scale multi-agent systems that detects and mitigates undesirable emergent norms, ensuring system compliance and stability.

Contribution

The paper presents AAF, a novel framework that records interaction provenance, detects norm violations, attributes responsibility, and applies interventions to maintain system integrity.

Findings

AAF reduces norm violations in simulations by 11.9% median

Detects violations with median delay of 71 steps

Achieves 97% attribution accuracy at 10% Byzantine rate

Abstract

Large-scale networked multi-agent systems increasingly underpin critical infrastructure, yet their collective behavior can drift toward undesirable emergent norms such as collusion, resource hoarding, and implicit unfairness. We present the Adaptive Accountability Framework (AAF), an end-to-end runtime layer that (i) records cryptographically verifiable interaction provenance, (ii) detects distributional change points in streaming traces, (iii) attributes responsibility via a causal influence graph, and (iv) applies cost-bounded interventions-reward shaping and targeted policy patching-to steer the system back toward compliant behavior. We establish a bounded-compromise guarantee: if the expected cost of intervention exceeds an adversary's expected payoff, the long-run fraction of compromised interactions converges to a value strictly below one. We evaluate AAF in a large-scale factorial simulation suite (87,480 runs across two tasks; up to 100 agents plus a 500-agent scaling sweep; full and partial observability; Byzantine rates up to 10%; 10 seeds per regime). Across 324 regimes, AAF lowers the executed compromise ratio relative to a Proximal Policy Optimization baseline in 96% of regimes (median relative reduction 11.9%) while preserving social welfare (median change 0.4%). Under adversarial injections, AAF detects norm violations with a median delay of 71 steps (interquartile range 39-177) and achieves a mean top-ranked attribution accuracy of 0.97 at 10% Byzantine rate.