Stories of AI chatbot users drifting from our shared reality are increasingly common, often described as cases of AI delusions, delusional spirals, or AI psychosis.
New research from Stanford University and the Human Line Project investigates the mechanisms behind these delusions, asking whether AI is making people more delusional or whether these chatbots are simply agreeing with delusional thinking.
You can binge more episodes of the Lab Notes podcast with science editor and presenter Jonathan Webb on the ABC Listen app (Australia). You'll find episodes on animal behaviour, human health, space exploration and so much more.
Get in touch with us:
[email protected]Featuring:
James Purtill, technology reporter
Further information:
The Dynamics of Delusion: Modeling Bidirectional False Belief Amplification in Human-Chatbot Dialogue
Characterizing Delusional Spirals through Human-LLM Chat Logs
This episode of Lab Notes was produced on the lands of the Gadigal, Ngunnawal and Ngambri people.