Julia Shaw

Shaw’s central claim is that “evil” is a continuum of ordinary human traits, not a separate category of monster. Psychopathy, sadism, narcissism, and Machiavellianism each run from low to subclinical or clinical levels, and social pressure can activate harmful behavior without hatred or pathology. For leaders and institutions, the risk model is therefore incentives plus conformity plus rationalization—not simply screening for “bad people.”

Generative AI is, in Shaw’s phrase, “the ultimate false memory machine,” because suggestive conversation can alter autobiographical memory within a single interaction. Her experiment produced confessions or apparent memories of invented crimes in 70% of one specific sample after three interviews, though she emphatically rejects treating that as a population rate. The product-safety implication is immediate: persistent personalization could magnify the risk, but even a 30-to-60-minute exchange can rewrite the version later recalled as truth.

The most reliable defense against memory contamination is contemporaneous evidence outside the brain. Shaw advises recording an important event immediately—even while drunk, high, or emotional—before discussing it with others, because “attention is the glue between reality and memory” and later recall retrieves the latest reconstructed version. AI systems could either contaminate that record or administer neutral cognitive interviews that preserve it.

Fraud scales by learning what a target wants to hear, making emotionally personalized AI a particularly potent attack surface. The Tinder Swindler combined love-bombing, luxury signals, an imagined shared future, and an urgent crisis until victims borrowed money; Shaw stresses that this works on ordinary people, not merely the gullible. “There’s a tailored version of fraud that could appeal to basically everybody if they have enough information about you.”

Criminal-justice spending is poorly aligned with repeat-risk if punishment is intended to make society safer. Homicide recidivism is only 1–3%, Shaw says, because most murders are unplanned fights over trivial triggers rather than serial predation; fraud, elder abuse, and sexual violence carry higher repeat risks. Long homicide sentences may satisfy retributive justice, but that is a different objective from prevention.

Corporate environmental crime emerges from normal organizational mechanics—benchmark pressure, collective deception, diffusion of responsibility, and fear of dissent. Volkswagen’s defeat-device scheme ran for 10 years and let some diesel cars emit nitrous oxides at up to 40 times the legal rate; clever engineers could know the health consequences and still conform. The investable risk is cultural and systemic: once one competitor appears to win by cheating, others infer that honest performance is impossible and misconduct cascades into litigation, consumer loss, and investor damage.

Shaw’s hopeful thesis is that the same technologies and institutions that distort reality can create auditable, science-backed accountability. Her company Spot, founded in 2017, applies a scripted cognitive interview to workplace reporting, while satellite monitoring, undercover investigators, Interpol, and existing environmental laws can make planetary crime measurable. The constraint is not a lack of tools but whether engineers include social-science knowledge and organizations protect the person willing to say, “This needs to stop.”

🔗 Original source & video: Julia Shaw: Criminal Psychology of Murder, Serial Killers, Memory & Sex | Lex Fridman Podcast #483