Belief-Sim: Towards Belief-Driven Simulation of Demographic Misinformation Susceptibility

2026-03-03

AI Summary

This work proposes BeliefSim, a novel framework that treats belief as the central driver of susceptibility to misinformation across diverse demographic groups. By integrating demographic attributes with psychological priors to construct individualized belief profiles, the framework leverages large language models under controlled prompting conditions to simulate susceptibility behaviors. Through post-training adaptation and counterfactual sensitivity analysis, BeliefSim achieves up to 92% simulation accuracy across multiple datasets, demonstrating that belief serves as a strong and effective prior for predicting susceptibility. This approach establishes a new paradigm for modeling misinformation propagation with both high precision and interpretability, highlighting the critical role of belief-driven mechanisms in shaping differential vulnerability among populations.

Technology Category

• Knowledge Representation and Reasoning: Reasoning with Beliefs
• Cognitive Modeling & Cognitive Systems: Simulating Human Behavior
• Application Domains: Misinformation & Fake News

Application Category

• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Social Networks and Social Media: Influence propagation, information diffusion, and the prediction on networks
• Responsible Web: Mitigating misinformation and disinformation

Abstract

Misinformation is a growing societal threat, and susceptibility to misinformative claims varies across demographic groups due to differences in underlying beliefs. As Large Language Models (LLMs) are increasingly used to simulate human behaviors, we investigate whether they can simulate demographic misinformation susceptibility, treating beliefs as a primary driving factor. We introduce BeliefSim, a simulation framework that constructs demographic belief profiles using psychology-informed taxonomies and survey priors. We study prompt-based conditioning and post-training adaptation, and conduct a multi-fold evaluation using: (i) susceptibility accuracy and (ii) counterfactual demographic sensitivity. Across both datasets and modeling strategies, we show that beliefs provide a strong prior for simulating misinformation susceptibility, with accuracy up to 92%.

Problem

• misinformation susceptibility
• demographic groups
• belief-driven simulation
• Large Language Models
• belief profiles

Innovation

• belief-driven simulation
• misinformation susceptibility
• demographic profiling
• large language models
• psychology-informed taxonomy