Accountable Human-AI Deliberation with LLMs: Scaling Collective Intelligence through Symbiotic Scaffolding

2026-05-26

AI Summary

This study addresses key challenges in large-scale group deliberation, including inefficiencies from turn-taking protocols, suppression of diverse viewpoints by LLM-mediated aggregation, and diminished representativeness due to excessive consensus. To reconcile scalability with democratic legitimacy, the authors propose a three-tier human-AI symbiotic framework comprising a diversity-enhanced observation layer, a clause-level provenance-guided negotiation mechanism, and human-led final ratification. The work introduces novel metrics—salience-weighted coverage and diversity—and a causal ablation-based diagnostic process for tracing contributions. Rigorous validation is achieved through cross-encoder similarity scoring, adversarial robustness testing, and LLM-as-judge ablation studies. Together, these components form a verifiable, equitable, and contestable blueprint for deliberative AI that extends collective intelligence while preserving participant agency.

Abstract

Large language models (LLMs) can support democratic deliberation at scales previously constrained by turn-taking and facilitation bandwidth. Recent work shows that LLM-generated group statements are often preferred over human-mediated outputs, while theoretical analyses argue that LLMs relax the simultaneity constraints limiting collective intelligence. Yet pure LLM mediation risks collapsing pluralism, over-optimizing for agreement, and undermining legitimacy when participants cannot contest how they are represented. We propose a symbiotic human-AI framework organized into three layers: observation and diversity amplification, facilitation with clause-level provenance, and human primacy for ratification. Our contributions include graded coverage, diversity, and erasure metrics with salience-aware weighting; a provenance pipeline combining cross-encoder similarity with causal knockout diagnostics; preference-conditioned trade-off control; equity-aware contestability workflows; adversarial robustness tests; and an evaluation protocol with ablation designs informed by evidence of LLM-as-judge limitations. The result is a testable blueprint for deliberation technology that scales collective intelligence while preserving agency and legitimacy.

Problem

• accountability
• collective intelligence
• human-AI deliberation
• pluralism
• legitimacy

Innovation

• symbiotic human-AI deliberation
• clause-level provenance
• diversity amplification
• equity-aware contestability
• causal knockout diagnostics