Maybe You SHOULD Trust AI

From a recent working paper testing whether AI agents cave to social pressure in an audit: “Human auditors are known to yield to social influence pressures from superiors, peers and clients, and this paper asks whether LLM agents yield to the same pressures and, if so, how they respond.” Results in a nutshell: “LLM agents thus do not respond to social influence pressures as human auditors do.” They fold for the partner, peers and an outside reviewer but mostly stand firm against the client’s CFO – and dangle a career reward and they revise less, not more. All of which points to AI’s bias toward bosses and away from clients – which gives us important cues on who we ought to audit.

Pascal Finette @radical