Harnessing Disagreement: Detecting Correlated Agreement Blindness in Multi-Agent Triage

2026-07-22

AI Summary

This study addresses the “correlation-induced consensus blind spot” in multi-agent arbitration, where model convergence can compromise safety monitoring in critical error regions. To mitigate this issue, the authors propose ARAT, a novel system that integrates inductive random forest and analogical k-nearest neighbor agents, augmented with a calibrated meta-model, a conservative coverage strategy, and a safety flag gating mechanism to foster constructive disagreement and alleviate consensus blind spots. Experimental results demonstrate that ARAT reduces the false negative rate from 4.80% to 1.70% on the UNSW-NB15 dataset and validates its efficacy in a cross-domain clinical readmission task. The findings further reveal that enhanced model capability may exacerbate error correlation, underscoring the necessity of diversity oriented toward constructive disagreement to strengthen system safety.

Abstract

Disagreement-triggered escalation can create a structural blind spot in multi-agent arbitration: as base learners improve, they tend to converge, weakening safety monitoring where correlated failures concentrate. We term this correlated agreement blindness and present ARAT (Arbitrated Reasoning Agents for Alarm Triage), a directed-star system combining an inductive Random Forest (RF) agent, an analogical case-based k-nearest neighbour (k-NN) agent, and a calibrated meta-model to mitigate this effect. On 82,332 holdout samples from the UNSW-NB15 network intrusion detection dataset, 57.2% of errors occur under agreement and 90.6% of dangerous under-predictions evade disagreement-based monitoring even after conservative override; ablation shows that strengthening base learners increases error correlation while reducing disagreement. ARAT reduces under-prediction relative to soft voting from 4.80% to 1.70% via conservative override (-2.6pp) and a safety-flag gate (-0.5pp), demonstrating architectural gains. Cross-dataset validation on clinical readmission supports these indicators, suggesting that diversification improves safety only when it generates productive disagreement rather than convergence. These results indicate that disagreement-triggered escalation can be blind to correlated failure, a risk that may intensify as agentic pipelines deploy increasingly capable, correlated models.

Problem

• correlated agreement blindness
• multi-agent triage
• disagreement-triggered escalation
• error correlation
• safety monitoring

Innovation

• correlated agreement blindness
• multi-agent arbitration
• ARAT
• productive disagreement
• safety monitoring