Cambridge DPhil student Ari Deller, who we supported with a scholarship in partnership with the Institute for Human Studies, recently published a paper in journal Philosophy & Technology on the epistemic risks posed by AI.

Ari argues that the real epistemic danger is less misinformation and more that super-persuasive models will saturate the environment with compelling but unreliable arguments – undermining confidence in even true beliefs formed via AI reasoning.

He also believes that by providing an instant, polished rationalization for whatever you happen to believe, LLMs risk making motivated reasoning frictionless.

Check it out here: