They gave lots of different descriptors.…They would just say, like, “They’re frowning.” Or they’d use one of these proverbs that say…it means his brow is dark, which obviously can translate as “He’s frowning.” They wouldn’t infer that that means that this person is angry.
To make sure the Trobrianders weren’t some kind of special case, Jarillo and Crivelli then traveled to Mozambique to study a group of isolated subsistence fishermen known as the Mwani. Once again, the results were dismal. The Mwani did marginally better than chance with the smiling faces, but they seemed baffled by sad faces and angry faces. Another group, led by Maria Gendron, traveled to the mountains of northwest Namibia to see whether the people there could accurately sort photographs into piles according to the emotional expression of the subject. They couldn’t.
Even historians have now gotten into the act. If you could go into a time machine and show the ancient Greeks and Romans pictures of modern-day people grinning broadly, would they interpret that expression the same way we do? Probably not. As classicist Mary Beard writes in her book, Laughter in Ancient Rome:
This is not to say that Romans never curled up the edges of their mouths in a formation that would look to us much like a smile; of course they did. But such curling did not mean very much in the range of significant social and cultural gestures in Rome. Conversely, other gestures, which would mean little to us, were much more heavily freighted with significance.
If you staged a screening of that Friends episode for the Trobriand Islanders, they would see Ross confronting Chandler and think Chandler was angry and Ross was scared. They would get the scene completely wrong. And if you threw a Friends premiere in ancient Rome for Cicero and the emperor and a bunch of their friends, they would look at the extravagant grimaces and contortions on the faces of the actors and think: What on earth?
5.
OK. So what about within a culture? If we limit ourselves to the developed world—and forget about outliers and ancient Rome—do the rules of transparency now work? No, they don’t.
Imagine the following scenario. You’re led down a long, narrow hallway into a dark room. There you sit and listen to a recording of a Franz Kafka short story, followed by a memory test on what you’ve just heard. You finish the test and step back into the corridor. But while you were listening to Kafka, a team has been hard at work. The corridor was actually made of temporary partitions. Now they’ve been moved to create a wide-open space. The room has bright-green walls. A single light bulb hangs from the ceiling, illuminating a bright red chair. And sitting in the chair is your best friend, looking solemn. You come out, thinking you’re going to be heading down the same narrow hallway, and BOOM—a room where a room isn’t supposed to be. And your friend, staring at you like a character in a horror film.
Would you be surprised? Of course you would. And what would your face look like? Well, you wouldn’t look the same as a Trobriand Islander would in that situation, nor a citizen of ancient Rome. But within our culture, in this time and place, what surprise looks like is well established. There’s a perfect example of it in that same Friends episode. Ross’s roommate, Joey, rushes into Monica’s apartment and discovers two of his best friends trying to kill each other, and his face tells you everything you need to know: AU 1 + 2 (eyebrows shooting up) plus AU 5 (eyes going wide) plus AU 25 + 26, which is your jaw dropping. You’d make the Joey face, right? Wrong.
Two German psychologists, Achim Schützwohl and Rainer Reisenzein, created this exact scenario and ran sixty people through it. On a scale of one to ten, those sixty rated their feelings of surprise, when they opened the door after their session with Kafka, at 8.14. They were stunned! And when asked, almost all of them were convinced that surprise was written all over their faces. But it wasn’t. Schützwohl and Reisenzein had a video camera in the corner, and they used it to code everyone’s expressions the same way Fugate had coded the Friends episode. In only 5 percent of the cases did they find wide eyes, shooting eyebrows, and dropped jaws. In 17 percent of the cases they found two of those expressions. In the rest they found some combination of nothing, a little something, and things—such as knitted eyebrows—that you wouldn’t necessarily associate with surprise at all.3
“The participants in all conditions grossly overestimated their surprise expressivity,” Schützwohl wrote. Why? They “inferred their likely facial expressions to the surprising event from…folk-psychological beliefs about emotion-face associations.” Folk psychology is the kind of crude psychology we glean from cultural sources such as sitcoms. But that is not the way things happen in real life. Transparency is a myth—an idea we’ve picked up from watching too much television and reading too many novels where the hero’s “jaw dropped with astonishment” or “eyes went wide with surprise.” Schützwohl went on: “The participants apparently reasoned that, since they felt surprised, and since surprise is associated with a characteristic facial display, they must have shown this display. In most cases, this inference was erroneous.”
I don’t think that this mistake—expecting what is happening on the outside to perfectly match what is going on inside—matters with our friends. Part of what it means to get to know someone is to come to understand how idiosyncratic their emotional expressions can be. My father was once in the shower in a vacation cottage that my parents had rented when he heard my mother scream. He came running to find a large young man with a knife to my mother’s throat. What did he do? Keep in mind that this is a seventy-year-old man, naked and dripping wet. He pointed at the assailant and said in a loud, clear voice: “Get out NOW.” And the man did.
On the inside, my father was terrified. The most precious thing in his life—his beloved wife of half a century—was being held at knifepoint. But I doubt very much that fear showed on his face. His eyes didn’t go wide with terror, and his voice didn’t jump an octave. If you knew my father, you would have seen him in other stressful situations, and you would have come to understand that the “frightened” face, for whatever reason, was simply not part of his repertoire. In crisis, he turned deadly calm. But if you didn’t know him, what would you have thought? Would you have concluded that he was cold? Unfeeling? When we confront a stranger, we have to substitute an idea—a stereotype—for direct experience. And that stereotype is wrong all too often.
By the way, do you know how the Trobrianders show surprise? When Crivelli showed up, he had a little Apple iPod, and the islanders gathered around in admiration. “They were approaching me. I was showing them.…They were freaking out, but they were not doing it like, ‘Gasp!’” He mimed a perfect AU 1 + 2 + 5. “No. They were doing this.” He made a noise with his tongue against his palate. “They were going click, click, click.”
6.
This is the explanation for the second of the puzzles, in Chapter Two, about why computers do a much better job than judges at making bail decisions. The computer can’t see the defendant. Judges can, and it seems logical that that extra bit of information ought to make them better decision-makers. Solomon, the New York State judge, could search the face of the person standing in front of him for evidence of mental illness—a glassy-eyed look, a troubled affect, aversion of the eyes. The defendant stands no farther than ten feet in front of him and Solomon has the chance to get a sense of the person he is evaluating. But all that extra information isn’t actually useful. Surprised people don’t necessarily look surprised. People who have emotional problems don’t always look like they have emotional problems.
Some years ago there was a famous case in Texas in which a young man named Patrick Dale Walker put a gun to his ex-girlfriend’s head—only to have the gun jam as he pulled the trigger. The judge in his case set bail at $1 million, then lowered it to $25,000 after Walker had spent four days in jail, on the grounds that this was long enough for him to “cool off.” Walker, the judge explained later, had nothing on his record, “not even a traffic ticket.” He was polite: “He was a real low-key, mild-mannered young man. The kid, fro
m what I understand, is a real smart kid. He was valedictorian of his class. He graduated from college. This was supposedly his first girlfriend.” Most important, according to the judge, Walker showed remorse.
The judge thought Walker was transparent. But what does “showed remorse” mean? Did he put on a sad face, cast his eyes down, and lower his head, the way he had seen people show remorse on a thousand television shows? And why do we think that if someone puts on a sad face, casts their eyes down, and lowers their head, then some kind of sea change has taken place in their heart? Life is not Friends. Seeing Walker didn’t help the judge. It hurt him. It allowed him to explain away the simple fact that Walker had put a gun to his girlfriend’s head and failed to kill her only because the gun misfired. Four months later, while out on bail, Walker shot his girlfriend to death.
Team Mullainathan writes,
Whatever these unobserved variables are that cause judges to deviate from the predictions—whether internal states, such as mood, or specific features of the case that are salient and over-weighted, such as the defendant’s appearance—they are not a source of private information so much as a source of mis-prediction. The unobservables create noise, not signal.
Translation: The advantage that the judge has over the computer isn’t actually an advantage.
Should we take the Mullainathan study to its logical conclusion? Should we hide the defendant from the judge? Maybe when a woman shows up in a courtroom wearing a niqab, the correct response isn’t to dismiss her case—it’s to require that everyone wear a veil. For that matter, it is also worth asking whether you should meet the babysitter in person before you hire her, or whether your employer did the right thing in scheduling a face-to-face interview before making you a job offer.
But of course we can’t turn our backs on the personal encounter, can we? The world doesn’t work if every meaningful transaction is rendered anonymous. I asked Judge Solomon that very question, and his answer is worth considering.
MG: What if you didn’t see the defendant? Would it make any difference?
Solomon: Would I prefer that?
MG: Would you prefer that?
Solomon: There’s a part of my brain that says I would prefer that, because then the hard decisions to put somebody in jail would feel less hard. But that’s not right.…You have a human being being taken into custody by the state, and the state has to justify why it’s taking liberty away from a human, right? But now I’ll think of them as a widget.
The transparency problem ends up in the same place as the default-to-truth problem. Our strategies for dealing with strangers are deeply flawed, but they are also socially necessary. We need the criminal-justice system and the hiring process and the selection of babysitters to be human. But the requirement of humanity means that we have to tolerate an enormous amount of error. That is the paradox of talking to strangers. We need to talk to them. But we’re terrible at it—and, as we’ll see in the next two chapters, we’re not always honest with one another about just how terrible at it we are.
Solomon: So while I guess there’s a sliver of my brain that’s saying, “Oh, yeah. Well, it’d be easier not to look,” I have the person looking at me and me looking at them. Having their family in the audience waving to me during the defense argument, you know, and he has three family members back here. It should be.…You should know that you’re impacting a person. It shouldn’t be taken lightly.
1 It was developed by legendary psychologist Paul Ekman, whom I wrote about in my second book, Blink. See the Notes for an explanation of how my views on Ekman’s work have evolved since then.
2 The plaintiff was Ginnah Muhammad. Her reply: “Well, first of all, I’m a practicing Muslim, and this is my way of life, and I believe in the Holy Koran, and God is first in my life. I don’t have a problem with taking my veil off if it’s a female judge, so I want to know, do you have a female that I could be in front of? Then I have no problem. But otherwise, I can’t follow that order.”
3 The 17 percent figure includes the three people (5 percent) who displayed all three expressions. Only seven people showed exactly two expressions. Also, although the vast majority of people believed they had expressed their surprise, one unusually self-aware person said he did not think his surprise had shown at all.
Chapter Seven
A (Short) Explanation of the
Amanda Knox Case
1.
On the night of November 1, 2007, Meredith Kercher was murdered by Rudy Guede. After a mountain of argumentation, speculation, and controversy, his guilt is a certainty. Guede was a shady character who had been hanging around the house in the Italian city of Perugia, where Kercher, a college student, was living during a year abroad. Guede had a criminal history. He admitted to being in Kercher’s house the night of her murder—and could give only the most implausible reasons for why. The crime scene was covered in his DNA. After her body was discovered, he immediately fled Italy for Germany.
But Rudy Guede was not the exclusive focus of the police investigation—nor anything more than an afterthought in the tsunami of media attention that followed the discovery of Kercher’s body. The focus was instead on Kercher’s roommate. Her name was Amanda Knox. She came home one morning and found blood in the bathroom. She and her boyfriend, Raffaele Sollecito, called the police. The police came and found Kercher dead in her bedroom; within hours they added Knox and Sollecito to their list of suspects. The crime, the police believed, was a drug- and alcohol-fueled sex game gone awry, featuring Guede, Sollecito, and Knox. The three were arrested, charged, convicted, and sent to prison—with every step of the way chronicled obsessively by the tabloid press.
“A murder always gets people going. Bit of intrigue. Bit of mystery. A whodunit,” British journalist Nick Pisa says in the documentary Amanda Knox—one of a vast library of books, academic essays, magazine articles, movies, and news shows spawned by the case. “And we have here this beautiful, picturesque hilltop town in the middle of Italy. It was a particularly gruesome murder. Throat slit, semi-naked, blood everywhere. I mean, what more do you want in a story?”
Other signature crime stories, such as the O. J. Simpson and JonBenét Ramsey cases, are just as enthralling when you rediscover them five or ten years later. The Amanda Knox case is not. It is completely inexplicable in hindsight. There was never any physical evidence linking either Knox or her boyfriend to the crime. Nor was there ever a plausible explanation for why Knox—an immature, sheltered, middle-class girl from Seattle—would be interested in engaging in murderous sex games with a troubled drifter she barely knew. The police investigation against her was revealed as shockingly inept. The analysis of the DNA evidence supposedly linking her and Sollecito to the crime was completely botched. Her prosecutor was wildly irresponsible, obsessed with fantasies about elaborate sex crimes. Yet it took a ruling by the Italian Supreme Court, eight years after the crime, for Knox to be finally declared innocent. Even then, many otherwise intelligent, thoughtful people disagreed. When Knox was freed from prison, a large angry crowd gathered in the Perugia town square to protest her release. The Amanda Knox case makes no sense.
I could give you a point-by-point analysis of what was wrong with the investigation of Kercher’s murder. It could easily be the length of this book. I could also refer you to some of the most comprehensive scholarly analyses of the investigation’s legal shortcomings, such as Peter Gill’s meticulous “Analysis and Implications of the Miscarriages of Justice of Amanda Knox and Raffaele Sollecito” in the July 2016 issue of the criminology journal Forensic Science International, which includes paragraphs like this:
The amplified DNA product in sample B was also subjected to capillary gel electrophoresis. The electrophoretic graph showed peaks that were below the reporting threshold and allele imbalance at most loci. I counted only 6 alleles that were above the reporting threshold. The electrophoretic graph showed a partial DNA profile that was claimed to match Meredith Kercher. Consequently, sample B was borderli
ne for interpretation.
But instead, let me give you the simplest and shortest of all possible Amanda Knox theories. Her case is about transparency. If you believe that the way a stranger looks and acts is a reliable clue to the way they feel—if you buy into the Friends fallacy—then you’re going to make mistakes. Amanda Knox was one of those mistakes.
2.
Let’s return, for a moment, to the theories of Tim Levine that I talked about in Chapter Three. Levine, as you will recall, set up a sting operation for college students. He gave them a trivia test to do. In the middle of it the instructor left the room, leaving the answers on her desk. Afterward, Levine interviewed the students and asked them point-blank whether they had cheated. Some lied. Some told the truth. Then he showed videos of those interviews to people and asked them if they could spot the students who were lying.
Social scientists have done versions of this kind of experiment for years. You have a “sender”—a subject—and a “judge,” and you measure how accurate the judge is at spotting the sender’s lies. What Levine discovered is what psychologists always find in these cases, which is that most of us aren’t very good at lie detection. On average, judges correctly identify liars 54 percent of the time—just slightly better than chance. This is true no matter who does the judging. Students are terrible. FBI agents are terrible. CIA officers are terrible. Lawyers are terrible. There may be a handful of “super-detectors” who beat the odds. But if there are, they are rare. Why?
The first answer is the one we talked about in Chapter Three. We’re truth-biased. For what turn out to be good reasons, we give people the benefit of the doubt and assume that the people we’re talking to are being honest. But Levine wasn’t satisfied with that explanation. The problem is clearly deeper than truth-default. In particular, he was struck by the finding that lies are most often detected only after the fact—weeks, months, sometimes years later.
Talking to Strangers Page 13