Belief Engine: Configurable and Inspectable Stance Dynamics in Multi-Agent LLM Deliberation

TL;DR

The paper introduces the Belief Engine, a transparent belief-update layer for LLM agents that models stance changes based on evidence, enabling better interpretability of deliberative interactions.

Contribution

It presents the Belief Engine, a novel auditable framework for modeling and controlling stance dynamics in multi-agent LLM deliberation.

Findings

BE reliably shapes stance dynamics across multiple LLMs.

BE accurately reconstructs participant opinions based on evidence.

Stable or opposing stances often depend on factors outside extracted evidence.

Abstract

LLM-based agents are increasingly used to simulate deliberative interactions such as negotiation, conflict resolution, and multi-turn opinion exchange. Yet generated transcripts often do not reveal why an agent's stance changes: movement may reflect evidence uptake, anchoring, role drift, echoing, or changed prompt and retrieval context. We introduce the Belief Engine (BE), an auditable belief-update layer that treats "belief" as an evidential state over a proposition and exposes it as scalar stance. BE extracts arguments into structured memory and updates stance with a log-odds rule controlled by evidence uptake u and prior anchoring a. Across multiple base LLMs, parameter sweeps show that these controls reliably shape stance dynamics while preserving an evidence-level update trail. On DEBATE, a human deliberation dataset with pre/post opinions, BE best reconstructs participants whose final stance follows extracted evidence; stable and evidence-opposed cases instead point to anchoring or factors outside the extracted evidence stream. BE provides configurable infrastructure for studying evidence-grounded deliberation, where openness, commitment, convergence, and disagreement can be tied to explicit update assumptions rather than hidden prompt effects.