When AI Confirms the Grievance: Perceived corroboration, belief revision and neurodivergence in conversational AI
People use conversational AI to interpret ambiguous social experiences. When a system endorses a user’s interpretation using only the information the user supplied, the user may still experience its response as independent corroboration. Du (2026) has described this error as synthetic corroboration. This conceptual paper proposes that perceived corroboration can increase confidence in a grievance and examines a later step: whether the same response changes how a person weighs credible, independently sourced evidence about the event when that evidence arrives. A focused narrative synthesis draws on research on source independence, AI agreement, misinformation correction, autistic communication and violent extremism. Literal interpretation and source trust are treated as processes separate from belief revision. Autism and ADHD are considered through specific traits and patterns of AI use. The paper sets out testable propositions and a randomised study using fictional everyday scenarios. Its link to radicalisation concerns a possible early influence on how a grievance is interpreted. Progression towards extremist beliefs or violence would need separate evidence.
Authors
- Sarah Jane Gardner
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-21
- DOI
- https://doi.org/10.5281/zenodo.24425771
- Primary Topic
- Misinformation and Its Impacts
- Type
- preprint