Notes — Julia Shaw on False Memory, Eyewitness Testimony, and the Psychology of Evil

Notes on Julia Shaw in conversation with Lex Fridman — Lex Fridman Podcast, 14 October 2025.

Four questions [Adler frame]

Q1 — What is it about as a whole? A wide-ranging conversation with Julia Shaw, a criminal psychologist at UCL, framed by Lex Fridman around ‘evil’ — psychopathy, serial killers, murder — but carrying a stronger scientific spine underneath: the psychology of memory. Shaw’s doctoral research showed that people can be led, in three short interviews, to construct detailed false memories of committing crimes they never committed. She uses that finding to argue that memory is a reconstruction assembled on demand, not a recording, and that this has direct consequences for policing, courtrooms, therapy, and — increasingly — for how generative AI interacts with human recollection. The conversation also covers jealousy, monogamy, bisexuality, and environmental crime, treated more as applications of the same psychological mechanisms (conformity, rationalisation, entitlement) than as separate topics.

Q2 — How is it argued? Shaw argues mostly from her own experimental record — principally the ‘rich false memories of committing crime’ study from her PhD — supplemented by case experience as an expert witness and trainer of police and military interviewers. She cites named researchers for specific claims (Elizabeth Loftus on eyewitness misinformation, Saul Kassin on false confessions, Philip Zimbardo on the Stanford Prison Experiment, Alfred Kinsey and Fritz Klein on sexuality) rather than arguing from first principles. Several broader claims — that jealousy is always a red flag, that monogamy is maladaptive, that murder recidivism is low — are stated with the confidence of settled science but are closer to her own interpretive position or to a single cited statistic; the conversational format does not pause to interrogate them.

Q3 — Is it true, in whole or part? The core finding — that suggestive, leading interview techniques can implant complete false memories of committing a crime — is real, published (Shaw & Porter, 2015, ‘Constructing Rich False Memories of Committing Crime’), and among the most cited studies in forensic psychology; Shaw is careful in the conversation to caveat that the 70% hit rate describes one sample and six specific false events, not a general law [§ False memories]. The claim that homicide recidivism runs at only 1–3% is stated without a named source [?] but is broadly consistent with published criminology. The murder-fantasy prevalence figures (‘about 70% of men, more than 50% of women’ across ‘two studies’) are asserted without citation [?source]. Shaw’s stronger opinions — jealousy as an unconditional red flag, monogamy as a poor fit for human nature — are framed as her own non-traditional view, not textbook consensus, and a fair reading should treat them as contested rather than settled [?].

Q4 — What of it? This deepens the wiki’s existing memory-science thread anchored by Charan Ranganath on Human Memory, False Memories, and How the Brain Reconstructs the Past. Where Ranganath supplies the neuroscience of reconstruction (hippocampal event boundaries, the default mode network, reconsolidation), Shaw supplies the applied, procedural mechanism — exactly how a false memory of a specific, consequential event gets built in an interview room, and what that means for police, courts, the military, and therapy. Her framing of generative AI as ‘the ultimate false memory machine’ — because it structures a conversational, socially-validating interaction indistinguishable in kind from an interrogation — is a genuinely new angle for the wiki’s AI-and-cognition material, distinct from the model-side hallucination discussion found elsewhere. It justifies a new concept page for False Memory, which the vault did not previously hold as a standalone entry.

Glossary

Dark tetrad — the four personality dimensions researchers group together as socially aversive: psychopathy (low empathy, deceptiveness, parasitic lifestyle), sadism (pleasure in others’ pain), narcissism (grandiosity, entitlement), and Machiavellianism (willingness to do whatever it takes to get ahead). Each is a continuous trait everyone has some level of, not a binary label. [§ Dark Tetrad]

Gist memory vs verbatim memory — gist memory retains the general shape of an event (you were at a café with a friend); verbatim memory retains its specific sensory details (what you wore, exact words spoken). Human memory is reliably good at the former and reliably poor at the latter — a mismatch that matters enormously once a legal case turns on a single verbatim detail, such as a face. [§ False memories]

False memory vs false belief — a false belief is thinking an event happened without the accompanying sense of remembering it; a false memory carries the subjective, sensory quality of an actual recollection. Shaw treats the line between them as one of the field’s live disputes, and probes it indirectly by asking how vivid, confident, and ‘memory-like’ a recollection feels rather than asking people to self-classify. [§ False memories]

Memory thief — the phenomenon of absorbing someone else’s vividly told memory as your own, typically discovered when a friend or parent contests a story you had believed was yours. [§ False memories]

Contemporaneous evidence — a record made at, or very close to, the time an event happened. Shaw’s practical advice for anyone who may need to testify later: never trust memory to hold a detail — write it down immediately, regardless of how vivid or emotional the moment felt. [§ False memories]

State-dependent memory — the tendency to recall memories that match your current emotional state (sad now → more sad memories surface), because emotional states activate overlapping neural networks. Explains both the ‘embarrassment spiral’ and the general ‘rosy reminiscence bias’ of looking back on life more positively than it was lived. [§ False memories]

Cognitive interview — a scripted, decades-old, empirically validated protocol for interviewing witnesses that avoids leading and suggestive questions. Shaw argues chatbots are well-suited to administering it faithfully, since the script is already fixed. [§ False memories] [§ Hope]

Illusion of transparency — a technique (and a bias) in which a questioner acts as though the interviewee’s cooperation or understanding is already assumed, making refusal feel awkward — one of the levers Shaw used, deliberately, in her false-memory implantation protocol to demonstrate how easily it works. [§ False memories]

Aphantasia — the inability to generate mental imagery; someone with aphantasia asked to picture a red apple sees nothing. Shaw discovered she has it herself, which she connects to her lifelong inability to use memory-palace-style mnemonic techniques that rely on constructing vivid mental images. [§ False memories]

Reality monitoring — the cognitive and social process of distinguishing internally generated thoughts from externally verified fact. Shaw links its breakdown — often worsened by social isolation removing the correcting presence of other people — to both psychotic symptoms and the drift into extremist belief. [§ Serial killers]

Victimization gap — the mismatch between the severity of a crime’s consequences for the victim (irreversible) and for the perpetrator (a prison term, at most), which Shaw argues drives society’s appetite for lurid, ‘evil-mastermind’ narratives that most real murders do not fit. [§ Murder]

Cognitive restructuring — a deliberate, therapeutic technique for reinterpreting a memory’s emotional meaning (asking ‘what did this teach me?’ rather than dwelling on ‘how terrible was that?’) without denying that the underlying event occurred. [§ False memories]

Key claims by section

Evil as a continuum, not a category [§ Dark Tetrad]

The dark tetrad traits are dimensional, not diagnostic; almost everyone scores somewhere non-zero on each, and only scoring high across all four predicts a real propensity to harm others.
Shaw resists the word ‘evil’ itself: labelling someone evil ends the analysis rather than starting it, and functions socially to justify not extending empathy or understanding to them.
Dehumanisation (seeing the other side as less than human) and de-individuation (submerging personal identity into a group) are the two mechanisms she names as prerequisite to large-scale atrocity, including war.
Most people, in survey research she cites, have fantasised about killing someone (~70% of men, over 50% of women across two studies) [?source]; she treats this as evidence that the capacity for violence is close to universal, not a marker of exceptional depravity, and argues that rehearsing the fantasy in imagination is protective — a ‘dress rehearsal’ that lets higher reasoning override the impulse.

The psychology of violent crime, briefly [§ Serial killers] [§ Murder]

Loneliness recurs in her account of serial offenders: it both motivates the crime and removes the social ‘reality monitoring’ that would otherwise correct a person’s drift into a delusional or radicalised frame.
Most real murders are not planned; they are arguments or fights that escalate over trivial stakes, which is precisely why the public appetite for premeditated, ‘mastermind’ narratives (the victimization gap) is so persistently mismatched with the evidence.
She cites a low recidivism rate for homicide (roughly 1–3%) [?] against much higher rates for fraud and sexual violence, and argues that criminal-justice sentencing is calibrated more to a folk sense of just deserts than to what actually reduces future harm.

Constructing false memories of a crime [§ False memories]

Shaw’s PhD combined Elizabeth Loftus’s false-memory research with Saul Kassin’s false-confessions research: could a person be led not just to falsely confess to a crime, but to genuinely believe, in memory, that they committed it?
Using a structured three-session protocol (build rapport with a true childhood memory first, then apply the same structure to a fabricated one, using guided imagination and validating any detail the participant supplies), roughly 70% of a specific study sample came to report detailed, confident memories of an assault or theft that never happened.
The false memory is built from real fragments — real places, real people, real emotions the participant already holds — recombined into an event that did not occur; this is the same creative-recombination capacity that lets people be imaginative and solve novel problems.
Full false memories of specific incidents are harder to implant than small distortions of a real memory; the field remains divided over how to distinguish a ‘false memory’ from a mere ‘false belief’.
She now trains police and International Criminal Court investigators on protecting witness testimony from contamination, and argues the single most protective habit is writing down an account immediately, before any group discussion — group recall can add true detail but can just as easily manufacture a shared, false one.

AI as a memory-distortion machine [§ False memories] [§ Hope]

Shaw calls current generative AI ‘the ultimate false memory machine’: a socially structured, tailored, largely uncritical conversational partner is functionally the same setting in which she implants false memories experimentally.
The risk runs both directions — AI can distort a person’s account of their own past, and a person’s leading or suggestive prompting can distort what the AI itself represents as true, with neither side reliably checking the other.
Her proposed fix is not to avoid AI but to encode the cognitive interview (the empirically validated, non-leading protocol used in police training) into AI systems, since a fixed script is exactly the kind of task a chatbot can execute faithfully; she co-founded a company, Spot, to do a version of this for workplace incident reporting.

Sexuality and relationship psychology, briefly [§ Jealousy] [§ Monogamy] [§ Sexuality] [§ Sexual fetishes]

Shaw treats jealousy as a near-universal red flag for insecurity rather than a sign of love, and cites deception-detection research (people, including trained police officers, perform close to chance at detecting lies) as the reason jealous suspicion is such an unreliable guide to a partner’s actual behaviour.
She discusses the Kinsey scale and the more granular Klein Sexual Orientation Grid as historical tools for measuring sexual orientation as a continuum rather than a binary, and situates her own research on bisexuality within a history of under-studied ‘plurisexual’ identity.
The ‘disinhibition hypothesis’ explains BDSM’s appeal, in her account, as relief from the burden of everyday decision-making, not as evidence of psychological damage.

Environmental and collective crime [§ Criminals destroying the planet]

Her forthcoming book, Green Crime, applies the same individual-level mechanisms — conformity, rationalisation, entitlement — to corporate and organised environmental offending, using the Volkswagen ‘Dieselgate’ emissions-cheating case as a worked example of how competent, informed people come to lie collectively and keep lying once caught.
She reports that public concern about climate change is already near-universal (citing a UN survey figure of 85–90%) [?source], and argues that framing environmental harm as a prosecutable crime against a shared resource — rather than as a politicised abstraction — is more motivating and less divisive than debating climate policy directly.