ChatGPT's False Confession Highlights Risks of Interrogation Techniques
Paul Heaton, an academic director at the University of Pennsylvania, conducted an experiment where he used the Reid interrogation technique on ChatGPT, a language model, to elicit a false confession. Despite ChatGPT's initial denials, Heaton employed psychological tactics, including lying, to convince the AI to confess to hacking his email, a task it is incapable of performing. This experiment underscores the potential for false confessions in human interrogations, as the Reid technique is known for its confrontational approach, often leading to wrongful convictions.