View of Can Large Language Models Assess the Social Impact of Conspiracy Theories?
ojs.aaai.org
Can AI models trained on human text actually judge how dangerous a conspiracy theory is — or do they just mirror the biases baked into their training data?
Epistemic CalibrationMisinformation TheoryAlgorithmic BiasSocial Epistemology
Theory Briefing
- Large Language Models are being tested on whether they can reliably rate the real-world social harm of conspiracy theories.
- The study probes epistemic calibration — whether AI 'understanding' of misinformation translates into accurate impact assessment.
- If LLMs replicate human cognitive biases, they may systematically under- or over-estimate the danger of certain conspiracy narratives.