Quote · Lex Fridman Podcast
#483 – Julia Shaw: Criminal Psychology of Murder, Serial Killers, Memory & Sex
Where this was said
False memories
At 2:23:57 · chapter starts 1:39:04
This is the episode's intellectual climax. Shaw's PhD research combined two previously separate fields: Elizabeth Loftus's false memory work and Saul Kassin's false confession research. The question: can you make someone believe they committed a crime that never happened? The methodology was painstaking — years to get through ethics — and involved selecting participants who had never experienced the target events (assaulting someone with a weapon, stealing, hitting someone), obtaining parental confirmation, building rapport with a true memory first, then introducing the false memory through guided imagination ('close your eyes and picture being 14, the police called your parents...'). Social reinforcement ('good job') was applied every time a participant generated a detail [1] — Julia Shaw "In just three interview sessions, Julia Shaw convinced 70% of participants they had assaulted someone or committed a crime as a teenager — …" 2:06:59 . The first participant said 'I remember a blue sky' and Shaw was grinning next to her supervisor because she knew it was working. Within three sessions, 70% of participants had developed complex false memories complete with specific people, places, and feelings — drawn from real elements of their lives but woven into events that never occurred [2] — Julia Shaw "What we've created with GenAI is basically the ultimate false memory machine. We have created a tailored experience of something that is mo…" 1:41:54 . Shaw is careful about what the 70% means: it doesn't mean 70% of all people can have false memories implanted; it means the phenomenon is real, possible, and alarmingly easy. She connects this to current concerns about AI — GenAI's conversational structure mirrors the exact conditions of a leading interrogation, making it potentially the most powerful false memory machine ever built.
Generative AI systems are structured as social conversations — the same format that Shaw has spent decades studying for its power to distort memory. When AI tells you what you want to hear, it's running the same psychological playbook as a leading police interrogation. The result is a machine that quietly rewrites your sense of reality.
In the 1940s, Alfred Kinsey — a biologist who studied wasps — sold out auditoriums talking about human sexuality. His Kinsey Scale showed that roughly half of men are not exclusively heterosexual, demolishing the binary model of sexual orientation decades before the modern conversation began.
Today twice as many women identify as bisexual compared to men, a reversal of the pattern seen in mid-20th-century sexuality research.
Alfred Kinsey's landmark postwar study found roughly half of men fell somewhere between exclusively heterosexual and exclusively homosexual on his 0–6 scale.
In just three interview sessions, Julia Shaw convinced 70% of participants they had assaulted someone or committed a crime as a teenager — events that never happened. The method used nothing more than leading questions, social reinforcement, and guided imagination. This is how innocent people end up in prison.
A study co-authored by Elizabeth Loftus found that AI-generated videos of yourself doing something make you significantly more likely to falsely believe the event actually happened.
