The word 'evil' ends the conversation

Julia Shaw opens with a pair of numbers that refuse to sit together. In two studies, she tells Lex Fridman, about 70 percent of men and more than 50 percent of women admitted they had fantasized about killing somebody. Recidivism among people imprisoned for homicide is only 1 to 3 percent. If the thought is that common and the second killing that rare, the useful question is no longer what makes a monster. It is why the rest of us never pick up the knife.

Shaw is a criminal psychologist whose books roam across psychopathy, violent crime, police interrogation, false-memory manipulation, deception, and human sexuality. *Evil* (titled *Making Evil* in the United Kingdom) takes up murder and sadism; *The Memory Illusion* takes up how recollection lies; *Bi* takes up bisexuality. Her new book, *Green Crime*, follows poachers, illegal gold miners, corporate frauds, and other people who destroy shared ground. She has trained police, lawyers, and staff at the International Criminal Court, and she co-hosts the BBC podcast *Bad People*.

The conversation that follows runs more than two and a half hours. It moves from the loneliness of serial killers to the kitchen-table problem of who did the chores, and it keeps circling a single claim: the capacity for terrible acts is not a switch in a few people. It is a set of traits, situations, and stories that ordinary minds already know how to run.

Psychology groups four traits under the dark tetrad: psychopathy, sadism, narcissism, and Machiavellianism. Sadism is pleasure in someone else's pain. Machiavellianism is doing whatever it takes to get ahead. Narcissism is taking too much pleasure in the self and ranking others below you. Psychopathy is often a thinness of empathy, plus deceptiveness and a parasitic way of living off other people. Shaw's first correction is that none of these is a binary. Each one is a continuum. A person can sit low on sadism and high on narcissism; most of us, herself included, land in the subclinical range, below a diagnosis and still on the same axes.

Early in *Evil* she poses the thought experiment of whether you would kill baby Hitler. The point is not the answer. It is whether you think people are born capable of extreme harm, or whether upbringing and circumstance grow that capacity over time. Psychologists who have gone through Hitler's life, she says, still argue "mad or bad." He showed features people associate with sadism and low empathy, but his early years did not already contain the later horrors. Her own answer is no. There is no straight line from infant to adult, and she does not think people are born evil.

She is nearly hostile to the word itself. Call someone evil, and you are saying this person is so unlike you that you need not bother understanding how they became capable of the act, because you never would. "I am good," the label implies. That split, she argues, is what makes people able to commit horrors against whoever they have marked as evil. The British title *Making Evil* nods at a Nietzsche line she likes: thinking evil is making evil. The attribute is a sticker we put on others. Nothing carries it as an essence.

Fridman notes that large wars almost always require both sides to cast the fight as good versus evil. Shaw adds a second mechanism: de-individuation, the moment you stop being a person and become an *us*. You need both tools. Empathy for the other side has to collapse, and the group has to swallow you, lending a cloak of justice even when you are in the wrong. In court, on sexual-abuse and murder trials, she hears "evil" constantly. What society actually wants, she says, is for the behavior to stop. The only path there is to find the psychological and social levers that put this person in that room. "We all have the capacity to kill people and murder people and do other terrible things. The question is why we don't."

Our sense of who is trustworthy, Shaw says, is not to be trusted. Creepiness was only recently given a working definition, and it is mostly this: the person is not following social norms. Behavior leaves the expected lane, and the alarm sounds. In one waiting-room study, people sat more chairs away from someone they knew had a severe mental illness. The research does not find that those patients are inherently more violent. People simply read them as weird, and they open physical and psychological distance. Attractiveness does not reliably save you; too far from the norm in either direction can make the assessment worse.

The same broken radar shows up in lie detection. Aldert Vrij, a leading researcher in the field, has shown across many studies that reliably spotting a lie is extremely hard. Police officers and investigative interviewers tend to be highly confident that experience lets them see through a witness or a suspect. In the lab they are no better than chance, and they still think they are. The cost is not only missed lies. An overconfident interviewer can lock onto an innocent person, press hard, and help produce a wrongful conviction. Shaw extends the point to ordinary life: a great deal of jealousy is an amateur lie-detection exercise run on uncertainty, not on knowledge.

Some evidence does suggest that people high in psychopathy are better at lying, including "faking good" at parole hearings — performing the safe prisoner until a committee makes the wrong call. Historically there has also been a fear that empathy-focused treatment teaches them to weaponize the appearance of care. Jennifer Skeem in California takes another route: programs that teach the rules of society and try to convince people that being pro-social is a more effective way to get what they want. Shaw's own method, when she works cases with many victims and visceral detail, is to treat the file as a puzzle. Emotional investment, she says, is the therapist's job. She is there with police and lawyers, looking for pattern. The risk she admits is the opposite of hardening: seeing beauty in every puzzle, like Prince Myshkin in Dostoevsky's *The Idiot*, until the moral landscape goes soft underfoot.

Asked for a disturbing case from *Bad People*, Shaw names Robert Pickton, the Canadian serial killer whose farm was nicknamed Piggy's Palace. She covered it because it bent her own career. As an undergraduate at Simon Fraser University she was taught by Stephen Hart, an expert witness at the Pickton trial, who kept students current on the case and still had gallows humor in the face of one of the worst people in Canadian history. That combination — kindness, seriousness, a joke at the edge of horror — is one reason she entered the field. Hart later came on the show. Elizabeth Loftus, a founder of false-memory research, appears in the same breath: she has educated courts on memory in highly controversial trials, including the Ghislaine Maxwell case, without claiming the right to vote guilty or innocent.

What kind of person feeds victims to pigs? Shaw's first word is loneliness. It is a pattern she sees in many serial killers. Loneliness does two jobs. It helps push the crimes, and it helps the person get away with them, because there is no social network doing what psychologists call reality monitoring: telling the difference between a thought in the head and a fact in the world. Schizophrenia is the clinical extreme. Command hallucinations tell a person to harm someone and never stop talking if they refuse; they are profoundly distressing, and they do raise the risk of violence. A family member or a therapist who asks "how can you tell this thought is real?" can pull the person back. Without that tether, they wander into whatever universe the mind has built. Radicalization, Shaw says, is the same mechanism without the diagnosis: drift toward a group divorced from evidence, grow more extreme, and lose the rope that would haul you out.

Stack the dark tetrad on top of that broken tether and the picture completes. Low empathy, high sadism, a belief that your own goals license the means. Jeffrey Dahmer, she says, seemed to be trying to assemble the perfect partner by killing people and sewing a new version together. "That's so sad," is the sentence she reaches for, not only "how awful." Of course it is atrocious. She still wants the sympathy, because the extreme is a form of something she has felt, and the gap between her and him is a pile of other factors. The core, she insists, is in everyone. History agrees in the ugly direction: neighbors who called each other friends have turned one another in at the start of wars. It does not take much, once friends have been recast as enemies.

That is why she thinks murder fantasies can be adaptive. Walking through the worst thing in your head is a dress rehearsal for the life you do not want. You picture the group at your side, the person across from you, the aftermath, and higher reasoning usually vetoes the animal flash. The line is rumination, especially about a specific person; that is a risk factor, and she would send that person to a psychologist. A fleeting thought on a single day she calls healthy. Philip Zimbardo's Stanford prison study has been torn apart in the years since, but Shaw still takes the demonstration: random assignment to guard or inmate rewrote how people saw one another in days. Zimbardo later pushed what he called the heroic imagination — rehearsing the moment you step in. The Kitty Genovese murder was long treated as proof that humans walk past dying strangers. Later bystander work did not back that story. Most of the time, Shaw says, bystanders do intervene. What freezes people is an already-gathered crowd: you read the room, no one has moved, so it must not be real. The wish not to stand out in a bad way is what often kills heroism.

If fantasies are so common, why do people actually kill? Television, Shaw says, has the motives upside down. Producers arrive with myths she refuses — above all the idea that everyone who does something terrible must have had a terrible childhood. What murder looks like most of the time is boring. A fight gets out of control. Someone owed four dollars. Someone stole a bike. A bad decision in the moment, an overreaction, no plan, no psychopath sharpening knives for months. Audiences dislike that story because of what she calls the victimization gap: the victim loses a life, the perpetrator at best loses freedom, and people want extreme consequences to have extreme causes. Intimate-partner homicide is among the top four situations for murder and is equally unglamorous, so it is told less often.

Very few people who kill wanted to do it in that instance, let alone again. That is why homicide recidivism is so low, and why Shaw thinks the justice system ranks risk badly. Fraud, elder abuse, and sexual violence carry much higher chances of happening again. Long sentences for murderers often satisfy an eye-for-eye sense of justice more than they make the street safer. She does not support the death penalty, though she can follow the rationalization. Forgiveness, she says, belongs to victims' families. Some want the harshest punishment, which she finds understandable. Others want restorative justice: an apology, an explanation, and — in the grimly common case of one teenage boy killing another — a chance for the second child not to have his life stolen too. If the only goal were deterrence, locking up people who have already killed would not be the obvious move.

Fridman asks about an episode of *Bad People* on incels. Shaw's account is about entitlement. Many crimes, not only this subculture, run on the feeling that a certain life was owed — the Disney version, the white picket fence — and that taking or destroying something is a legitimate reply to not receiving it. Nobody is entitled to a good life, she says, even if she wants rights and freedoms arranged so people can pursue the life they actually want. Sold an impossible script, angry, and sitting in a group that agrees, a person can become dangerous. The internet, they both note, is very good at reinforcing that worst self.

Shaw interviewed Cecilia, who was taken in by the so-called Tinder Swindler. The method was ordinary and devastating. A man posed as a wealthy traveler, matched quickly, and love-bombed: constant messages, big emotion, a private jet, the future narrated on the third date — kids in the garden, you are the only one. He flew her toward Paris, expensive restaurants, the props of wealth. Then came the emergency. He was being followed, under attack; he needed money for flights. She took out loans. Her family gave money. When the money ran out, the story collapsed. Shaw's point is not that Cecilia was stupid. Fraud works because the other person knows what you want to hear and says it. Given enough information, she thinks, there is a tailored version that would work on almost anyone. Love is an especially efficient door. Large language models, Fridman notes, are getting better at building that tailored story; Shaw agrees that emotion is the easy way in.

Shame makes it worse. These cases get told as "she was so gullible." Shaw's counter is that she may have been a decent person who wanted to believe someone could love her. Victims often become cynical and close the capacity to be vulnerable; both speakers hope they do not. The same pattern, they agree, shows up inside non-criminal relationships as coercive control — a partner, often a man along old gender lines, taking the finances and turning shared money into an "allowance." Jealousy can be used the same way.

Shaw's view of jealousy is not the field's default. She is not a jealousy researcher, but she thinks it is basically always a red flag. It means the jealous person is not secure, either because the relationship is wrong or because they are insecure in themselves. She does not read it as love or as mate-guarding. She reads it as control and possession. Not all jealousy becomes violence, but intimate-partner violence almost always has jealousy as a precursor, often of imagined betrayals, running on that same broken lie detector. Research also finds that most people are bad at monogamy: many have cheated on a partner, if not the current one. Shaw flips the usual moral. Monogamy, she argues, is a social construct that sets a lot of people up to fail. She is polyamorous. She is not campaigning for everyone to live that way. She wants it on the menu, and she wants the conversation early — how you want to structure a life, money, exclusivity — because it is much harder to raise years in.

Fridman pushes back in a way she does not dismiss. For him, choosing one person is like finding a meal he likes and staying with it: an infinitely long project of understanding another human. Shaw's reply is that she wants people to feel they have a choice. All they currently see is monogamy. Honesty and communication, they agree, are the load-bearing walls; a persistent identity as "a jealous person" is the warning, not a single pang of being left out. Going into any structure, including polyamory, to fix something broken in yourself is a hard way to start. Bisexual people, her research finds, are more likely to end up in non-traditional relationships, in part because they are constantly asked to justify a sexuality that already refuses the one-gender script.

Shaw came out as bisexual in *Evil*, on purpose, while writing about visibility. Once you realize that people you already know and like belong to a group that had felt foreign or frightening, dehumanization gets harder. The book was also strategy. Publishing two scientific books first was a foot-in-the-door: *Bi* was the first mainstream book on bisexuality, and she does not think she could have placed it, or been taken seriously as a scientist, if it had come first. She later founded a bisexual research group because the literature was hard to find. Search "bisexual" on Google Scholar and you hit biology about two sexes; researchers often use *plurisexual* instead. The right picture, she says, is not that she is attracted to most people. It is that gender is irrelevant to the pool, the way a heterosexual person's pool is limited by gender and still excludes most humans.

The dominant misunderstanding in the research is that bisexuality is a phase, a stepping stone on the way to Gay Town. That stereotype hits men hardest. Today twice as many women as men identify as bisexual; historically it was the other way around. After World War II, Alfred Kinsey — a biologist who had been working on gall wasps — was asked to teach a course on human sexual behavior, discovered that almost none of the students' questions had answers in the literature, and went out with assistants to interview thousands of Americans in the 1940s. *Sexual Behavior in the Human Male* was a bestseller for a year. He sold out halls; they piped radio into the next room. He built the Kinsey Scale from 0 to 6, exclusive heterosexual to exclusive homosexual, because a binary was not useful. About half of men and a quarter of women landed somewhere in the middle. Shaw notes the homosocial setting of wartime, prisons, and other same-gender worlds as part of the context. By the 1970s the public story had flipped toward women.

Social pressure, not inner truth, still shapes who says the word. Gay men tell bi men "I was bi once." Lesbians often read bi women as performing for men and destined to go back. Both directions can be hostile; for men the pressure to pick gay or straight is stronger, especially if their life already runs through queer parties and apps. The costs are not abstract. Bisexual people are more likely to self-harm, to be victims of sexual violence, to be isolated, and to be stalked. Shaw's explanation is simple: they are the least plugged into a community. They do not fully belong in the straight world and often do not feel fully held by the queer one.

Fritz Klein, a therapist in the 1970s and 1980s, built a finer grid on top of Kinsey: attraction, behavior, fantasy, social preference, lifestyle, and self-identification, each scored for past, present, and ideal. Shaw thinks everyone should fill it out. Friends of hers have come out past-straight, present-straight, ideal-bi. She looked at her own lifestyle column and thought it was painfully straight. Klein called clients uncomfortable in their orientation "troubled bisexuals"; she extends the idea to troubled straight, gay, or asexual people. The work is finding the sticking point. She dislikes the phrase "be honest with yourself." It is a high bar, and men in particular have used it as a weapon — you are not honest, you are actually gay. Maybe, she says, you just have not had the framework or the words. Sexuality is treated as something found in the teens or early twenties and then locked. People change on every other dimension. She wants room to reappraise.

Shaw put kinks in a book about evil because other people already had. Someone will ask her to explain a killing and then add, as if it were the key, that the person was into BDSM, or swinging, or some other nontraditional structure. Her reply is "so?" Unless a partner was the victim, it is not related to the crime. Broader society still reads kink as a defect of character, which she finds wild given the research: a great many people have at least one fetish, and BDSM — choking, restraints, consensual degradation — is the most common. Pornography can make it look ubiquitous, which it is not; marginalizing it as a tiny perversion is also wrong. The distribution sits in between. The rule is to ask, to talk, and to have a safe word. The range she names runs from pup play to blood play. People often feel deep shame about what they want, and shame is what makes them think they are bad. That is the same chain that still villainizes queer lives in much of the world. Destigmatizing, for her, is how you stop people rotting in self-disgust.

Why would someone want to be told what to do in bed when they spend all day refusing that in the office? Research on BDSM keeps returning to the disinhibition hypothesis. After a day of millions of decisions, a scene in which you make none is a release. The dominant role has a twin freedom: you do not have to walk on eggshells; the rules are clear. The paradox, Shaw says, is that you are free because of the fiction. You are playacting, and it still touches something psychologically real — the same reason people love novels, which let you borrow another mind for a while. Visibility is the civic version of that honesty. She wrote *Bi* in part because she is "constantly terrified" that rights will be stripped back. There is no straight line of acceptance. Fan mail arrived from countries where the book would count as contraband; one foreign publisher bought the rights and then wrote, after the law changed, that they could no longer print it as part of a "gay agenda." She takes a little pride in a banned book and a lot of sadness in what the ban means.

Coming out in her thirties was, on balance, freeing. As a young woman already in public she was being sexualized anyway; naming the identity, she found, often disarmed the worse comments and recruited allies. Advice to people still figuring it out: read widely, and if you think in constructs and numbers, try the Klein grid. The path into this whole career, she adds, was not a straight line either. She had an art portfolio ready — painting, surrealism, messing with the sense of reality — until her grandfather warned her off an artist's life. She applied for social sciences instead, already knowing it would be psychology, because her father has paranoid schizophrenia. Growing up with someone who was seeing and hearing different things in real time made "what is real" the question of her life: in morals, in memory, in crime, in perception. Criminal psychologists, she found, had gallows humor and still had nice lives. That was the vibe she wanted.

Shaw's PhD was the first study to braid two literatures: Saul Kassin's work on false confessions and Elizabeth Loftus's work on false memories. The question was whether you could get people not only to confess to a crime that never occurred, but to remember it. The short answer is yes, with leading and suggestive interview techniques. She has since trained police, lawyers, and people at the International Criminal Court who handle hundreds of witnesses and the problem of collective memory. The applied question is how to keep a memory as close to original as possible. That is also, she says, why the jump to generative AI is not far. We now talk to models as if they were interviewers. You get confabulation from both sides, and a neglected third thing: the in-between, where each party is pulling the other off the facts. "What we've created with Gen AI is basically the ultimate false memory machine."

False memories, she insists, are a feature of a healthy brain, not a glitch. She is talking about autobiographical memory — lived experience, not trivia. Basically every such memory is false; the question is only how false. Gist memory, the scientists' term, is usually good enough. You remember that you saw a friend, which university you attended, whether an event was happy or awful. Verbatim detail — what you wore, drank, said — is where people fall apart, and criminal justice is built on exactly that layer. If the detail is a face, the trouble is obvious. There are also complete fabrications: events that never happened. Many people discover them while talking to parents or friends. "That happened to me, not you." Research calls the person a memory thief. Someone else's vivid story was imagined, felt real, and filed under the wrong life.

The method was designed to show what is possible, which meant doing what she tells police not to do. Participants signed up for a study of emotional childhood memories and knew their parents had been contacted. Shaw collected real teenage details — where they lived, who their friends were — and confirmed they had never been through the target events. Each session began with about twenty minutes on a true memory, often a skiing accident, using a structured interview that she would later mirror. Then came the planted story: at fourteen they had been in contact with the police for assaulting someone with a weapon, usually something as ordinary as a rock, or for stealing, or for hitting someone. When they said they did not remember, she offered a detailed account and an "imagination exercise," using the culturally familiar and scientifically disputed idea of hidden or repressed memory. Eyes closed. The first trivial detail she ever heard, grinning beside a worried head of department, was "I remember a blue sky." Social reinforcement followed every addition. The specific people and places had to come from the participant; false memories are real fragments woven into a shape that never occurred. After three interviews, 70 percent of the sample had a full false memory and a confession.

They cut the study short. Ethics had been told to expect about a 13 percent hit rate. Shaw is careful with the number. It does not mean 70 percent of humanity can be given this particular false crime. It means that in this sample, with these six events and this interviewer, more people than not developed a complex memory. She suspects almost everyone is prone to some version of the trick. The field still argues about whether a false *belief* ("I think it happened") can be separated from a false *memory* ("I remember it"). As a social psychologist she only has what people say. She asks about texture: confidence, sensory feel, whether it resembles their other memories. Other researchers think the line is cleaner. She thinks it is almost impossible to draw.

A team including Loftus published work this year, Shaw says, showing that if you turn a person's own photos into AI video, they become more likely to believe the depicted events happened as the model invented them — and to remember them with high confidence. A large replication of her crime-implantation study is underway at the University of Maastricht. Fridman asks whether a government could mass-gaslight a population this way. "Yes. That is definitely already happening." She is not nostalgic for a lost age of truth. For most of history people could not even write; the scientific method, reliable catalogs, and instant fact-checking are late inventions. The new discomfort is that we got used to feeling we could reach unmodified information, and that feeling is getting harder to justify.

Defense starts with accepting how the mind works. Then: do not trust it. Assume you will forget, no matter how important, emotional, or certain you feel. Prospective memory — "I will remember" — fails. Contemporaneous evidence, the closer to the event the better, is the high-quality version. A popular myth says that if you were drunk, high, or flooded, you should sleep it off and recall later. Current memory advice, she says, is the opposite. Write it in the moment. You can write it again in the morning. At least you have an original. She trained German military agents who met warlords abroad and could not put a phone on the table. Their habit was to debrief as a team before writing notes. That felt like remembering more. It did not mean the extras were right. They now write first. Her colleague Annelies Vredeveldt finds that once each person has a private version, remembering together can add complementary detail — one partner holds what happened, the other holds when. The same process, without notes, can efficiently build a castle that never stood.

As an expert witness she wants version history. When a witness's story is categorically different three years later, the question is whether they were shy the first time, under pressure, genuinely remembering more — or whether hypnotherapy or other suggestive treatment, which can mirror her implantation protocol, has made the evidence too dirty to parse. How do you choose between the first account and the later one? "Receipts." Original notes and external evidence. Without them, two people telling opposite stories become a battle of confidence. She is pointed about therapies that hunt retroactively for a single childhood incident that explains who you are now. That, she thinks, is a wild theory of the brain. A therapist's job is to help someone feel better in the present; her job is evidentiary quality. Those roles can pull in opposite directions. Once a memory has been through a closed-eye, imagine-the-details, confirm-and-repeat loop, she can no longer tell which parts the mind supplied and which the treatment produced.

Fridman describes a private experiment: dwell on the decent parts of old relationships and the past warms; dwell on the fights and it darkens. Shaw says the essence is right, with a correction. State-dependent memory means you more easily retrieve what was stored in the state you are in now. Sadness lights the sadness network and offers more of the same. Embarrassment is the classroom example: one public stumble, and for six hours the brain queues every other humiliation you have ever had — spreading activation along neighboring synapses you cannot shut off. Happy people remember more happy scenes. Most of us, unless depressed, look back through a rosy reminiscence bias. What actually shifts is not a filter that keeps only objectively good events. It is the interpretation. Growing up with a father who has paranoid schizophrenia was complicated at the time. In hindsight she files it as a net positive: it defined her life and gave her a view of the mind she would not otherwise have. That, she says, is what good therapy should do — leave the facts, change the emotional content.

The deliberate version has a name: cognitive restructuring, changing an aspect of a memory on purpose so you can live better. Instead of "how terrible was that?" she asks what she learned, what it gave her that other people do not have, what it revealed about who her friends are. She thinks that move is part of resilience and should be taught more than hanging in the negatives. Daniel Kahneman's holiday thought experiment sits nearby. Offered a perfect tropical vacation you will not remember, many people decline. A large share of life's pleasure, the result suggests, happens in the retelling. Shaw is not against the phone photo. She is against using the phone as a substitute for attention. "Attention is the glue between reality and memory." Outsource the noticing, and later you will look at the picture and ask where it was. Autosuggestion can implant a false memory without another person in the room: you read or hear a story, wonder whether something like it happened to you, picture it, and eventually file it as yours. Neuralink-style tools, she thinks, would be doing a faster version of a thing language already does.

Her favorite relationship finding is almost comic. Ask a couple who does most of the housework. One says 60 percent, the other says 50 percent, and the total exceeds 100. People overclaim their own contribution and shrink their partner's, on many fronts. The unglamorous fix is to say the task out loud: I just took out the bins; I just booked the hotel. It feels like bragging. It is how the other person's inner star chart ever gets the data. We assume they remember because we do. They may not have perceived it at all.

Memory can be trained, she says, if you avoid the Sudoku trap. Doing one puzzle type over and over caps out because it is the same move. She likes short, varied games — NeuroNation, developed with researchers including people at the University of Bonn — because the brain has to stay flexible. Short and many, not one thing hard: the opposite of expertise. Names defeat both her and Fridman. The usual trick is to enlarge the network with a rhyme, a picture, an absurd hook, and to actually listen when the name is said. She recently learned why the picture-based tricks fail for her. She has aphantasia: no mental imagery. Close your eyes and picture a red apple; she sees black. Some people are hyperphantasic, apple in high detail; some get a gray outline; she gets nothing. Memory palaces sound absurd to her. People with aphantasia, research finds, care less about childhood memories because they cannot see them. She has an untested hunch that analytical people may be over-represented. A lot of her friends, it turns out, have it too. Most do not know until a conversation like this one.

*Green Crime* is an experiment: can criminal psychology be pointed at people destroying shared ground? Shaw's analogy is a house. If someone walked in, set fires, poisoned the pipes, and dumped garbage, anger would be the correct response. That is happening at planetary scale. Her case study is Dieselgate. For ten years Volkswagen built diesel cars with a defeat device that made laboratory tests look clean while nitrogen oxides on the road ran, in some cases, at forty times the legal limit. There is no healthy lower bound of those gases for the human lung; they are tied to asthma, premature death, and immediate harm. The people who built the software were clever emissions engineers. They knew what the exhaust did. After they were caught, they kept lying; the truth took about a decade. Conformity, rationalization ("the market works this way; if I don't, someone else will"), and diluted agency ("it was the company's decision") are the same tools she has been describing all night. Honest firms see a rival's impossible numbers, infer cheating, and start cheating. A race to the bottom ends in lawsuits, furious investors, furious customers.

Environmental crime is layered in a way street murder is not. There are bosses, middlemen, and people on the ground with guns, boats, saws, pipes. There is the surreal extra layer: a fishing vessel on Interpol's criminal register for ten years that still finds an insurer. Violent crime, in her usual caseload, is small-scale, often an argument, often a perpetrator who is also vulnerable — substance use, mental illness, sleeping rough. Here some of the smartest people in the world are still running frauds and shell companies. A United Nations climate survey she cites finds that something like 85 to 90 percent of people think about the climate crisis regularly; most do it every day. The idea that only a minority cares is an illusion. The feelings are eco-anxiety, anger, sadness, grief. She wants those feelings turned into purpose. Framing the problem as crime has a side benefit: nobody debates whether a rainforest just burned. At a European Space Agency meeting she watched scientists present satellites pointed at Earth, cataloging change in many wavelengths. A lot of this, she says, is visible from space. The remaining work is how to chase, punish, and prevent.

Fridman brings up *Ordinary Men*: Nazi execution squads in Eastern Europe, men told they could opt out, most of whom stayed, apparently without hatred in the heart — just enough social pressure not to be the one who chickens out. Shaw answers with the emissions engineer who has a child with asthma and a struggling grandmother and still ships the dirty car. In these stories there is usually a moment when someone inside or outside says it has to stop. That is the heroic imagination again, and the bystander problem, applied to a planet. Does she have empathy for those criminals? "I have empathy for everybody." Has it ever failed? She pauses on sexual slavery: women, often trafficked, held in a confined space and forced into repeated acts while the person running it sits outside. She is not calling it the worst crime. For her mind there is a door marked you cannot go there. Fridman says cases involving children are that door for him. The admission matters because her whole brief is evil-empathy, and she will not pretend the brief has no edge.

What gives her hope is layered. People study darkness at all. Technology, used on purpose, can carry scientific methods into ordinary life: she shares the prompt she uses so a model will run a proper cognitive interview, the scripted protocol British police already read from a page, which a chatbot follows more faithfully than a tired human. In 2017 she co-founded Spot with Phil Libin, the founder of Evernote, and two partners, Dillon and Daniel. It grew from the existential hangover of her talks. She would explain how memory lies, watch the room deflate, and leave. Spot is a workplace reporting tool at talktospot.com: a chatbot that administers the cognitive interview, captures an original account of an important emotional event, and can send a report to an employer. Insurers, medical companies, and — her favorite credential — the Bar Council in England use it. People have reported someone leaving bleach in a washing machine. Fridman asks whether the same idea could hold a life's important events, good and bad, for models that currently have no such memory. In Spot's product, no; in private life she is interested, including grief bots that can represent you or someone you love. She jokes that she is waiting for the email. He says he knows people at DeepMind. "Hit me up." Most of the evil in the world, Fridman says in closing, quoting T. S. Eliot, is done by people with good intentions. That is Shaw's argument in one line. Once you are sure you are the good one, you stop checking.