PAVE: A Cognitive Architecture for Legitimate Violation in Generative Agent Societies

2026-05-19

AI Summary

This work addresses the challenge of enabling generative agents to appropriately violate limited rules in emergency scenarios such as fire evacuations. The authors propose PAVE, a cognitive architecture comprising four modules: Perception, Appraisal, Verdict, and Emulation. PAVE introduces an explicit legality judgment mechanism that integrates situational severity with individual personality thresholds, allowing agents to perform bounded, recoverable, and authority-compliant rule violations only when necessary. Implemented with large language models and evaluated in the Voville traffic environment, PAVE outperforms baseline approaches in legality assessment, authority adherence, violation scope control, and state recovery. Human evaluations confirm that its decisions are perceived as more reasonable and trustworthy. Ablation studies further validate the critical role of the legality gating mechanism in guiding appropriate agent behavior under duress.

Abstract

Generative agents based on large language models reproduce believable human behavior in cooperative settings, but how they should reason in situations where rule-breaking may be required, such as fire evacuation or authority-supervised emergency, remains poorly characterized. We propose PAVE (Perception, Assessment, Verdict, Emulation), a novel four-module cognitive architecture that addresses this gap end to end: (i) Perception extracts a structured context with explicit authority distance, peer behaviors, and severity-tagged situational cues; (ii) Assessment scores the context along five scalars including an explicit legitimacy judgment that checks necessity, proportionality, and absence of alternatives; (iii) Verdict decides to comply or violate under a hard legitimacy gate, with a per-agent threshold elicited from the persona; (iv) Emulation enacts the verdict and scopes the violation to the rule the trigger justifies. We instantiate PAVE in Voville, a tile-based traffic environment forked from Smallville, and evaluate across three scenarios, four LLM backbones, and a focused ablation. PAVE agents satisfy four properties simultaneously: legitimate violation (only when a trigger justifies it), authority deference (officer instructions override even high legitimacy), bounded scope (violations confined to the targeted rule), and recovery (baseline restored once the trigger ends). PAVE agents make more structured and interpretable decisions than vanilla across all four properties, and human evaluators rate them as more plausible. Ablating the legitimacy gate reproduces vanilla-like failures. We release Voville, the PAVE prompts and code, and the evaluation pipeline.

Problem

• generative agents
• rule violation
• legitimacy
• emergency reasoning
• cognitive architecture

Innovation

• cognitive architecture
• legitimate violation
• generative agents
• large language models
• rule-breaking reasoning