AI-based Credibility Assessment

Automated systems that use machine learning and AI to evaluate the credibility, veracity, or trustworthiness of online content. These systems generate credibility signals or predictions about whether content is true, false, misleading, or lacks sufficient evidence. AI credibility assessments are increasingly deployed on social media platforms to help users judge misinformation.

AI credibility indicators can operate at different levels (post-level, account-level, source-level) and employ various feature types: textual features (language patterns, claims), network features (user and content propagation patterns), and external signals (fact-checking databases, authoritative sources).

A key challenge is the interaction between AI predictions and human judgment: users may over-rely on weak AI signals, under-rely on strong ones, or use AI predictions selectively to rationalize prior beliefs.

Key papers

• One AI Signal, Many Human Judgments: A Bayesian Cascade Analysis of AI-based Credibility Indicators in Online Information Spread — extends Bayesian cascade models to show how AI credibility signals reshape sequential user judgments
• A Survey of Fake News: Fundamental Theories, Detection Methods, and Opportunities — comprehensive survey including credibility assessment methods

Related topics

• Misinformation and fake news detection (broader — includes non-AI methods)
• Content moderation (application domain)
• Human-AI Interaction (how users interact with AI signals)
• Information Cascades (how AI signals propagate through populations)