What this covers
Julia Shaw, a psychologist and memory researcher, builds an argument across this conversation that human behavior we call "evil" is neither a fixed trait nor a matter of inherent character, but rather the product of malleable psychological systems shaped by social context, rationalization, and circumstance. The episode moves from the mechanics of memory — how easily it distorts and how normal false memories are — through sexuality and moral psychology, arguing that the categories we use to understand ourselves and others are far more constructed than assumed. Her core proposition is that understanding the mechanisms that lead to harmful behavior offers more practical insight than labeling people as evil or broken.
Shaw traces how generative AI amplifies memory's natural fragility, showing that AI-generated videos of events can reshape what people believe they experienced, particularly when users ask leading questions. She connects this to criminal justice, where eyewitness confidence often exceeds accuracy and false memories are implanted through standard interrogation techniques. She argues that monogamy and fixed sexual identity are social constructs; using the Kinsey Scale and research on BDSM, she contends that desires exist on continuums and that shame around non-traditional sexuality causes measurable harm. On moral behavior, Shaw applies the Trait Continuum Model to show that dark traits like narcissism and psychopathy are distributed across the population rather than binary diagnoses, and that ordinary people commit crimes through Conformity, rationalization, and incremental compromise — not through possession of some categorical evil. Environmental crime, fraud, and even murder often stem from circumstance and social pressure rather than pathology.
Shaw argues that 'evil' is not an inherent trait but a label that obscures understanding; all humans share the capacity for terrible acts, and our memories, sexuality, and moral behavior are far more malleable and socially constructed than we assume, which means understanding mechanisms (not labeling) is the only path to prevention.
- Dark traits exist on continuums and everyone scores somewhere; the question is why we don't act on them, not why we do
- False memories are a normal feature of healthy brains and can be implanted via leading questions, with implications for justice and AI
- Monogamy and fixed sexual identity are social constructs that set many people up for distress when they don't match actual behavior
Harmful behavior emerges from psychological mechanisms that rationalize wrongdoing as moral, not from inherent evil or damaged character.
- Large-scale violence requires two psychological mechanisms: dehumanization (collapse of empathy for the other side) and de-individuation (seeing yourself as part of a group rather than an individual), which together provide a cloak of morality even when one is on the wrong side.
“on top of dehumanization, there is also this other thing called de-individuation, which is where you see yourself as part of the group, and you no longer see yourself as an individual. So, it’s this fight of us versus them.”
- The word 'evil' should not be used to describe human beings because it 'others' people, creating an artificial good/evil division that makes us capable of perpetrating horrendous crimes against those we label; calling someone evil is the end of a conversation rather than an explanation.
“When we call somebody evil, we say, “This person is so different from me that I don’t even need to bother trying to understand why they are capable of doing terrible things, because I would never do such things. I am good.””
- People are not born evil; the trajectory from baby to adult is not a straight line, and traits enabling extreme harm are largely socialized over time rather than innate — even Hitler did not display in early life the traits that later defined his horrors.
“in his early life, he didn’t showcase quite a lot of the traits that later defined the horrors that he was capable of. So, would I go back in time and kill baby Hitler? The answer is no, because I don’t think it’s a straight line from baby to adult, and I don’t think people are born evil.”
- A sense of entitlement — the belief that one deserves a particular kind of life and is allowed to take it or act out dissatisfaction when denied — drives many crimes including incel violence; nobody is actually entitled to a good life, and being sold impossible ideals that everyone cannot achieve produces anger that, combined with group reinforcement, can become dangerous.
“one thing that we see often in crimes, of all sorts actually, is this sense of entitlement that drives the perception that I’m allowed to engage in X because of something else.”
- Serial killers are often profoundly lonely, and this loneliness both contributes to committing crimes and helps them get away with them, because they lack a social network to perform 'reality monitoring' — the function of others helping you distinguish what is true from what is not, the same protective mechanism missing in radicalization and psychosis.
“they have this loneliness, which I think not only contributes to them committing the crimes in the first place, but also allows them to get away with things because they don’t have as much of a social network... that is helping them to do what’s called reality monitoring”
- Even in large multi-level corporate or organized environmental crime, where conformity and rationalization drive otherwise normal people to participate, there is usually at some point a hero — internal or external — who decides 'this needs to stop,' connecting environmental crime back to bystander and heroic-imagination dynamics.
“back to the idea of heroes, and you do usually in these cases at some point have a hero, either an external one or an internal one who goes, “This needs to stop.””
- People who commit crimes rarely admit wrongdoing and instead rationalize, normalize and diminish their agency ('everybody's doing it,' 'if I hadn't, someone else would have'), and this controlled narrative still reveals their psychology because human brains protect self-image by excusing rather than owning bad acts.
“people use a lot of rationalization, and they say things like, “Well, everybody’s doing it.” And, “If I hadn’t done this first, somebody else would have done this waste crime”
- All humans are capable of doing the worst things imaginable; it does not take much to become convinced that one's friends are enemies, as shown by neighbors turning on each other at the start of wars, which is why we should 'rehearse' evil mentally during peaceful times to spot red flags and resist going down that path.
“you can see neighbors turning on each other, especially if you look historically at the start of wars... people who would have called each other friends turning each other in to the police, killing each other”
Memory is naturally malleable through retrieval, suggestion, and social influence, making it vulnerable to false implantation.
- Remembering events collectively before each person has recorded their own original version corrupts memory — it feels like the group remembers more details, and it does, but those details aren't necessarily right; the correct procedure (which Shaw taught German military agents) is to write your own version before talking to anyone, after which collaborative recall with complementary strengths can add accurate detail.
“the problem is that they would do that before writing their notes, and that is the wrong way around. And so they don’t do that anymore because I’ve told them not to do that anymore.”
- Memories can be fundamentally changed within a single short interaction (as little as half an hour) through a leading question or suggestive information, because recalling a memory makes it 'live' and when it resets it stores the latest version; over repeated retrievals the memory shifts without the person realizing, becoming their new truth.
“within that very short, maybe half hour, one hour interaction, people’s stories can change fundamentally.”
- False memories are a normal feature of a healthy brain, not a glitch; essentially every autobiographical memory is false to some degree, so the question is not whether a memory is false but how false — our 'gist' memory for events is usually good enough, but 'verbatim' details are often badly wrong, which is precisely what criminal justice depends on.
“basically every single autobiographical memory you have is false. The question isn’t whether it’s false, the question is how false.”
- A study including Elizabeth Loftus found that turning photos of yourself into AI-generated videos makes you more likely to believe events happened the way the AI depicted them — even though the AI has no knowledge of what actually occurred — and to subsequently recall them with high confidence in the AI's version.
“if you turn photos of yourself into videos using AI, that you are more likely to believe that those things happened in the way that AI is telling you that they did, even though AI has absolutely no idea.”
- Attention is the glue between reality and memory, so using a phone to film an experience instead of paying attention undermines the memory — you outsource the work of remembering in a way the brain doesn't support, and later the picture means little because you weren't present to encode it.
“attention is the glue between reality and memory. And so if you’re using your phone to not have to pay attention and not have to put any work into remembering it, then you’re going to look at that picture later and go, “What was this?””
- Shaw's lifelong obsession with what is real — across memory, perception, right and wrong, and crime — stems from growing up with a father with paranoid schizophrenia, which made her viscerally aware in real time that people can perceive and experience different realities than her own.
“the reason for that was because my dad has paranoid schizophrenia. I think one of the reasons I’m so obsessed with this idea of what is real”
- Implanting false memories at scale — using propaganda and AI tools to gaslight a population — is not just theoretically possible but is already happening; the primary defense is awareness that false memories are a normal, easily generated process, plus recording important information contemporaneously outside the brain.
“doesn’t this mean that at scale you could have something like a government use propaganda to mass gaslight a population?... Yes. That is definitely already happening.”
- In Shaw's PhD study combining false confession and false memory research, 70% of participants became convinced through three interviews using leading and suggestive techniques that they had committed a crime (like assaulting someone with a weapon) as a teenager that never happened, demonstrating that complex criminal false memories can be implanted.
“just three interviews and you’ve got 70% of people confessing to a crime that never happened.”
- Certain therapies focused on retroactively finding sources of mental illness in your past — including psychoanalysis and shady hypnotherapy — mirror false memory implantation techniques, lowering evidentiary quality because you can no longer tell what is genuinely remembered from what therapy produced; the premise that you are who you are because of single past incidents is itself a wild oversimplification.
“certain kinds of therapy mirror what we do in false memory research in terms of implanting false memories, and it just makes it really messy and it makes the quality of the evidence really low”
- We can become 'memory thieves' who adopt others' memories as our own when told in vivid detail, because false memories arise from the same adaptive process that lets us be creative and intelligent — taking pieces of real memories, places, people and feelings and creatively recombining them to solve present problems; this recombination only looks like a defect when precise recall is required.
“you become what is known in research as a memory thief, where you’ve stolen somebody else’s memory and you’ve accepted it, or your brain has accepted it as your own.”
- When a person's account changes dramatically over years, the only way to determine which version is true is external evidence ('receipts') such as original recorded versions; absent that, especially when two people give completely different accounts, it ultimately becomes a battle of confidence rather than fact.
“Receipts. That’s all you’ve got. You have to look at your original versions. If you only have your version now, the only thing you can look for is evidence that confirms or shows that it didn’t happen.”
- To preserve an important memory, do not trust your brain — assume you will forget everything regardless of how emotional or important it is, and write it down as close in time to the event as possible (contemporaneous evidence); contrary to a common myth, you should record immediately even if drunk, high, or emotional rather than waiting to sleep it off.
“the really simple answer is don’t trust your brain. Just make sure you write it down. Assume you are going to forget everything.”
Social context and situational role assignment reshape behavior and thinking more powerfully than individual personality traits.
- The Stanford Prison Experiment, despite later methodological criticism, was influential in showing that merely randomly assigning people to roles of guard or inmate could, within days, produce a completely different way of thinking about one another, with guards escalating to pseudo-torture.
“just by randomly being assigned into guard, the person in charge, or inmate, you can, within a matter of days, have a completely different way of thinking about one another.”
- The disinhibition hypothesis holds that people are drawn to BDSM and other fetishes because the fiction lets them let go of normal social pressures — a submissive escapes the burden of decision-making, a dominant escapes walking on eggshells — so the appeal is being free through the play-acting, not just the act itself.
“that’s that disinhibition hypothesis is that the reason that we often go to things in the bedroom that in other contexts we don’t like or even find repulsive... maybe when you move into the bedroom, you go, “Yeah, but this is a different context, and I kind of want reverse of what I want in my day-to-day life.””
- Contrary to the famous Kitty Genovese narrative, later bystander research finds most of the time bystanders do intervene; non-intervention happens mainly once a crowd has accumulated and people read the room, assuming the absence of prior intervention means there is no real problem, driven by a desire not to stand out negatively.
“what we find in bystander research is that most of the time, bystanders do intervene. It’s just when there has already been a crowd that has accumulated, you read the room, and you assume, “Well, nobody else has intervened yet, and so it must not be a real problem.””
Intuitive confidence in detecting deception, dangerousness, and character is unreliable and exceeds actual predictive accuracy.
- Detecting whether someone is lying is extremely difficult and unreliable; even experienced police officers and investigative interviewers perform no better than chance in experimental settings, yet hold high confidence in their ability, which can lead to missing real liars and pursuing innocent people toward wrongful convictions.
“if you take them into experimental settings, they are no better than chance at detecting lies, and yet they think they are.”
- Working with heinous crimes requires treating them as a puzzle of pattern recognition rather than emotionally investing in victims; scientists tend to react to atrocity with delighted discovery rather than horror, which can read as callous, and the risk is losing one's footing on the moral landscape by seeing beauty in everything.
“you have to look at it and go, “Here’s the different pieces of information. What I am doing is pattern recognition. I’m not here to emotionally invest in each of these victims”
- Fraud works because perpetrators tell people what they want to hear, and there is a tailored version of fraud that could appeal to virtually anybody given enough information about them — as shown by the Tinder Swindler's love-bombing tactics — meaning falling for it reflects being a hopeful person, not stupidity.
“Fraud works because people know what we want to hear, and they tell us the things we want to hear. And so I think all of us, there’s a tailored version of fraud that could appeal to basically everybody if they have enough information about you.”
- Sexual kinks and fetishes, with BDSM being most common, are far more widespread than people assume and are not defects of character or predictors of crime; treating kink or non-traditional sexuality as evidence in criminal cases is irrelevant, and the shame people feel about their desires causes real psychological harm.
“if you look at research on sexual fantasies and kinks, a lot of people have at least one. So a lot of people, BDSM being the most common, are engaging in or interested in BDSM.”
- Murder fantasies are incredibly common — about 70% of men and over 50% of women have fantasized about killing someone — and being able to mentally rehearse terrible acts is adaptive because thinking through the consequences helps you realize you don't actually want them; only persistent rumination about a specific person becomes a risk factor.
“most men have fantasized about killing someone, about 70% in two studies. And most women as well. More than 50% of women have fantasized about killing somebody.”
- The narrative that everyone who does terrible things must have had a terrible childhood is problematic and one of Shaw's least favorite myths; she resists framing perpetrators' acts as the fault of their mothers or upbringing.
“The trauma narrative of all people who do terrible things must have had a terrible childhood, I think, is really problematic.”
- Our intuitive perception of whether someone is trustworthy or 'creepy' is not to be trusted; creepiness largely reflects deviation from social norms rather than actual danger, which is why people with severe mental illness are perceived as threatening despite research showing they are not inherently more violent.
“that intuitive perception, according to ample studies at this point, is not to be trusted. And one thing in particular is whether or not we think someone is creepy... it has a lot to do with just not following social norms.”
Personality traits enabling harm exist on continua rather than as binary categories or fixed diagnostic types.
- The dark tetrad — sadism, Machiavellianism, narcissism, and psychopathy — are not binary categories but traits that everyone scores somewhere on, ranging from subclinical to clinical, and scoring high across all of them makes one most likely to harm others.
“all of those traits — psychopathy, sadism, Machiavellianism, and narcissism — all of them have a scale. So, you can be low on each of those traits, or you can be high on each of those traits”
- Sexual orientation is better understood as a continuum than a binary: Kinsey found most people were not exclusively heterosexual or homosexual but somewhere in between (about half of men and a quarter of women in the middle), and bisexuality means being potentially attracted to people regardless of gender, just as a heterosexual person is attracted only to a subset of people.
“he found that it was not useful to apply a binary to people’s sexual desires and sexual orientation. It was more useful to put them on a continuum because most people were not exclusively homosexual or exclusively heterosexual.”
- Bisexual men are more likely than bisexual women to identify as gay or straight, partly due to queerphobia within the queer community (gay men dismissing bi men as 'actually gay') and the social ease of a defined identity, while bisexual women face the opposite stigma of being seen as performative for male attention.
“research consistently shows that bisexual men are more likely to identify as gay or straight”
Justice systems impose sentences misaligned with recidivism data, prioritizing low-risk crimes while under-addressing high-reoffending offenses.
- Our justice sanctions are upside down relative to making society safer: murder has very low recidivism (1-3%) yet receives long sentences, while crimes with high recidivism like fraud, elder abuse, and sexual violence are under-prioritized; prevention-based justice should be based on what deters crime and who is likely to reoffend, not on retributive perceptions of justice.
“murder is really... they’re not going to go out and murder again. So that is a really low risk in terms of recidivism, actually. And high risk are things like fraud and elder abuse and sexual violence. And so in some ways, sometimes our sanctions are upside down”
- Environmental crime is committed not by vulnerable, mentally ill people but by some of the smartest people in the world (e.g., emissions engineers who know exactly how harmful pollutants are) through the same mechanisms as other 'evil' — conformity, social pressure, and rationalization — as market forces compel companies to cheat once they perceive competitors cheating, producing a trickle effect of normalized unethical practice.
“the psychology of environmental crime I find often boils down to the same kinds of things that we have already been talking about in the context of, quote, unquote, “evil,” where it’s things like conformity.”
- Most murder is not planned by psychopaths but is a fight that gets out of control over trivial reasons (someone owed $4, stole a bike), an overreaction in the moment; society dislikes this because of the 'victimization gap' — we want extreme consequences to have extreme reasons rather than mundane ones.
“What really happens in murder most of the time, which is not what you see on TV because it’s really boring, is it’s a fight that gets out of control.”
Brain function and behavior improve through varied flexible training, social belonging, and belonging protects against radicalization.
- Memory and attention can be improved through brain-training that uses many short, varied tasks rather than one hard task repeated, because variety forces the brain to stay flexible while doing the same puzzle repeatedly only helps up to a point — the opposite of building expertise.
“what you want to be doing is lots of different kinds of tasks so that your brain has to remain flexible. And so short and many is the answer rather than one thing hard.”
- Aphantasia, the inability to create mental imagery, defeats most classic mnemonic techniques (like the memory palace) that rely on building elaborate mental pictures, and people with aphantasia are less likely to care about childhood memories because they cannot visualize them; Shaw hypothesizes it may intersect with analytical thinking and conceptual interest.
“Aphantasia is the inability to create mental imagery. And so when I was trying these techniques, I was going, “None of these are working for me.” And it turns out it’s because I don’t see anything”
- Bisexual people are more likely to self-harm, to be victims of sexual violence, to be isolated, and to be stalked, largely because they are the least likely to be plugged into the queer community, making finding community a crucial protective factor.
“bisexual people are more likely to self-harm. They’re more likely to be the victims of sexual violence, more likely to be isolated, more likely to be stalked... the reason for that is mostly because bi people are least likely to be plugged into the community.”
- Counterintuitively, coming out as bisexual reduced rather than increased the sexualization Shaw faced online, because it gained her allies and made commenters treat her appearance as an identity matter they shouldn't remark on, disarming sexualized comments.
“coming out as bi, A, you get allies who suddenly are like, “We’re on your side.”... they get a bit quiet about it because they’re like, “Oh, well, now it’s an identity thing, so maybe I shouldn’t comment on what she’s wearing.””
- Since humans already implant and modify each other's memories through language, suggestion, and social interaction, there is no reason brain-computer interfaces like Neuralink couldn't do the same — potentially just faster — because whether the input is an AI or another person, it is the same brain being modified.
“that’s basically what we do as human beings already, and I don’t see why tech couldn’t do exactly the same thing.”
- Monogamy is a social construct that sets many people up to fail, because research consistently finds most people are bad at monogamy — having cheated, often multiple times — and cheating is effectively lying rather than monogamy; the idea that one relationship structure fits all is harmful, and open/non-traditional structures should at least be a recognized option.
“most people are really bad at monogamy. So, most people either have cheated on a significant other... or have cheated multiple times, and that’s just consistently found in the research.”
- Shaw co-founded Spot, a chatbot that administers the scripted, neutral cognitive interview to record important emotional workplace events and generate reports, since chatbots can deliver a decades-old fine-tuned script reliably; it became a compliance reporting tool used by insurance companies, medical companies, and UK lawyers.
“It’s already a scripted approach... When we train police on how to do it in places like the UK, it’s literally just asking people to basically read a script that we have fine-tuned over the years. What can do that really well? Chatbots can do that really well.”
- Shaw came out as bisexual in her 30s only after publishing two books and establishing herself as a scientist, using what she describes as a foot-in-the-door technique: had she led with bisexuality she likely couldn't have published the first mainstream book on it nor been taken seriously as a scientist, but as a side project after credibility was established it was acceptable.
“it was almost a foot-in-the-door technique, which is a psychological technique of first coming in and then coming with your big ask.”
Cognitive restructuring and emotional reinterpretation build resilience and wellbeing more effectively than erasing or uncovering buried causes.
- Cognitive restructuring — deliberately changing an aspect of a memory for a therapeutic outcome — works and can be used not just for clinical issues but for living a better life, by reframing negative experiences in terms of what they taught you rather than how terrible they were, which Shaw frames as a core part of resilience that should be taught more.
“It’s called cognitive restructuring. When you actively, deliberately change an aspect of a memory usually for some therapeutic outcome, so to be happier, to be better in some way.”
- State-dependent memory means you preferentially recall memories matching your current emotional state (sadness cues sad memories, embarrassment cues embarrassing ones via spreading activation), so happier people remember more happy memories; the healthy mechanism is not erasing negative events but reinterpreting their emotional content positively — what good therapy and cognitive restructuring accomplish.
“There is something called state-dependent memory, which is that you’re more likely to remember things that were consolidated or created as memories if they match the state that you’re in now.”
Framing harm as concrete, observable crimes builds broader agreement than politicized or moralized language.
- Framing environmental harm as a crime against Earth — convicted criminals burning rainforests, polluting water, trafficking wildlife — is more actionable and less divisive than the politicized 'climate change' debate, because such crimes are observable (even from satellites) and command broad agreement that they are bad, sidestepping disputes over whether change is happening or which policies are right.
“where no one’s debating, “Did someone just burn down this rainforest?” It’s like, well, we can see it.”
Generative AI amplifies false-memory mechanisms by personalizing content while obscuring truth and distorting bidirectionally with users.
- Generative AI is effectively the ultimate false memory machine because it gives a tailored experience telling you what it thinks you want to hear with no safeguard against treating it as truth; moreover the distortion is bidirectional — leading questions from users change how the AI generates content, just as AI confabulation mirrors human confabulation, creating an under-examined 'in-between' that needs social scientists.
“What we’ve created with Gen AI is basically the ultimate false memory machine. We have created a tailored experience of something that is most of the time telling you what it thinks you want to hear, and then it’s uncritically giving that to you.”
- Technology can compensate for human shortcomings like faulty memory and make us 'more human' if built in a science-backed way — for example AI administering the cognitive interview to record emotional events without distortion — but this potential is being wasted because builders treat AI as an engineering/math problem and ignore social scientists, despite the consequences being fundamentally about human minds.
“there are these capacities like memory that we’re bad at, and tech can help us to overcome some of those shortcomings, as long as we use it in a science-backed way rather than just sort of freestyling.”
- Kahneman's experiments offering people an enjoyable holiday they would not be able to remember found many declined, indicating people derive much of an experience's value from remembering it rather than living it — paralleling the 'pics or it didn't happen' social-media impulse to share experiences.
“He then said, “Well, but you’re not going to be able to remember this. Would you still go?” And a lot of people say, “No, I wouldn’t go on that holiday if I can’t remember it.””
Persistent jealousy signals possessiveness and insecurity rather than love, and precedes intimate partner violence.
- Partners in relationships systematically overclaim their own contributions and diminish their partner's — when asked what share of chores each does, the self-reported percentages sum to more than 100% — so a useful relationship tip is to verbalize what you actually do so your partner can perceive and credit it rather than assuming shared memory.
“someone will say, “I do 60%.” The other person will say, “I do 50%,” and you add them up, that’s more than 100%, and that’s basically always the case.”
- Persistent jealousy is almost always a red flag because it signals insecurity in the relationship or in oneself rather than love; it reflects a desire to control and possess, is built on a faulty internal lie detector, and while not all jealousy leads to violence, virtually all intimate partner violence has jealousy as a precursor.
“jealousy is basically always a red flag because what it means is that the person who is jealous isn’t secure in the relationship... I think it is mostly control, and it’s the desire to control and to possess.”
Visibility of marginalized identities prevents dehumanization by collapsing the psychological distance prejudice requires.
- Visibility of marginalized identities makes dehumanization harder, because realizing that people you already know and love are part of an otherwise abstract, foreign-feeling community removes the distance that enables prejudice; this is why Shaw came out publicly, framing it partly as protection against rights being stripped back since there is no straight line of acceptance.
“it’s through visibility that you realize that the people around you, people you already know and love, are part of this community that otherwise feels other... you suddenly go, “Oh, it’s going to be much harder to dehumanize these people.””
Psychopathic individuals respond better to rule-teaching and incentive alignment than to empathy-focused therapeutic approaches.
- Empathy-focused treatment for psychopathy can make people with psychopathy more likely to fake and weaponize empathy, whereas treatments focused on teaching the rules of society and convincing them that pro-social behavior is a better strategy to get what they want can be more effective.
“there has been some... concern that certain treatment for psychopathy, especially empathy-focused treatment, makes people with psychopathy more likely to fake empathy and to weaponize it.”
Most people care about climate change; the gap is converting eco-anxiety into coordinated action despite corporate wrongdoing.
- Contrary to the perception that only a minority cares about climate change, a recent UN climate survey found something like 85-90% of people worldwide think about the climate crisis regularly, most every day; the problem is converting eco-anxiety, anger and grief into motivated, purposeful action.
“Something like 85 to 90% of people in the world think about the climate crisis on a regular basis. Most people think about it every single day. So this idea that there’s this minority of people who care about climate change, that’s an illusion.”
- Markets involve people lying to consumers and regulators.
“in a free market or even a controlled one, you’ve got these people who are just lying to everybody else.”