Truth-default theory

AI deception detection models show human-level accuracy but inconsistent bias

Researchers have developed seven Retrieval-Augmented Generation (RAG) models to detect deception, comparing their performance against baseline models using over 39,000 judgments across five deception datasets. The study…