Not Born Yesterday

by Hugo Mercier

  The potential benefits of following the majority are so sub-

  stantial that many nonhuman animals rely on this heuristic.24

  Baboons offer a perfect example. They travel in troops of sev-

  eral dozen members. The individuals forming the group must

  constantly make decisions about where to look for food next.

  To study their decision making, Ariana Strandburg- Peshkin and

  her colleagues outfitted the baboons from a troop with a GPS,

  al owing the researchers to closely track their movements.25

  Sometimes, the group starts splitting, with two subgroups

  72 ch ap t er 5

  Figure 2. “Bridge” strip from the webcomic xkcd by Randall Munroe.

  Source: xkcd .com.

  initiating a move in diff er ent directions. When this happens, the

  other baboons look at the numbers in each subgroup: the larger

  one subgroup is compared with the other, the more likely the

  baboons are to follow it.

  If baboons and other animals living in groups have an intui-

  tive grasp of the power of majority rules, it would be bizarre if

  humans, who rely much more on social information than ba-

  boons, completely neglected this rich source of insight.26

  The study mentioned earlier with the ninety- nine assembly

  members, and others like it, show that people do not grasp the

  power of majority rules when the question is framed abstractly,

  with the number of people forming the majority being repre-

  sented by a figure or percentage. Diff er ent results are obtained

  when we can see the opinions of individual people. In one of the

  cleanest studies on the topic, psychologist Thomas Morgan and

  his colleagues had participants solve a variety of tasks, for in-

  stance, deciding whether two shapes seen from diff er ent angles

  are the same or diff er ent, tasks that were built to be difficult

  enough that participants would not be certain of their own an-

  swers.27 The participants were given the individual answers of a

  number of other participants (or so they thought; in fact, the an-

  swers were made up by the experimenters). In this context,

  w h o k n o w s b e s t ? 73

  people were perfectly able to follow majority rules rationally:

  they were more likely to change their minds when a given answer

  was supported by a larger group, and when it was more consen-

  sual (holding group size constant). Such a pattern has been

  found in many other experiments, and it emerges in children dur-

  ing the preschool years.28

  We have an intuitive grasp of the power of majority rules. Yet,

  when the question is framed in more explicit, abstract terms, we

  display no such comprehension. A way of making sense of this

  apparent contradiction is to introduce the concept of an evolu-

  tionarily valid cue.29 A cue is evolutionarily valid if it was pre sent

  and reliable in the relevant period of our evolution. For instance,

  our ancestors were better off staying away from rotting meat; rot-

  ting meat generates ammonia; ammonia is an evolutionarily

  valid cue that food should be avoided; this is why we find its odor

  so repulsive (think cat pee).

  Assuming that other primates, like the baboons, are able to

  follow majority rule, the number of individuals making a given

  decision has likely been a reliable cue for a very long time, long

  before the human line split from that of the chimpanzees. By con-

  trast, numbers, odds, and percentages are recent cultural inven-

  tions.30 For this reason, they do not constitute evolutionarily

  valid cues. While it is pos si ble to react appropriately to such

  cues—as when we go through the equation of the Condorcet

  jury theorem— this takes dedicated, explicit learning.

  We find the same pattern when it comes to considering

  whose opinion influenced whom.31 The Condorcet jury the-

  orem fully applies if the voters have formed their opinions in-

  de pen dently of each other. If ninety- eight members of our

  imaginary assembly mindlessly follow the ninety- ninth mem-

  ber, the opinion of the assembly is only as good as that of the

  ninety- ninth member.

  74 ch ap t er 5

  Abstract cues related to the dependence between opinions,

  such as correlation coefficients, are simply ignored.32 This makes

  sense given that correlation coefficients are definitively not an

  evolutionarily valid cue. By contrast, when Helena Miton and I

  introduced evolutionarily valid cues, participants took them into

  account.33 Our participants were told of three friends who had

  all recommended a given restaurant. However, they had done so

  because a common fourth friend had told them the restaurant

  was great. In this case, participants knew to treat the opinion of

  these three friends as if it were the opinion of a single friend.

  Other experiments suggest that even four- year- olds are able to

  take some forms of dependence between opinions into

  account.34

  Resisting the Pull of the Majority

  When cues are evolutionarily valid, people weigh a number of

  them to decide how valuable the majority opinion is: size of the

  majority in relative terms (the degree of consensus) and abso-

  lute terms (group size), competence of the members of the ma-

  jority, and degree of de pen dency between their opinions.35 But

  how do they weigh all this in relation to their own prior state of

  mind? As we saw in chapter 1, it is widely believed that when

  these cues converge, in par tic u lar when people are faced with a

  large consensual group, the force of the majority opinion be-

  comes nearly irresistible.

  Asch’s conformity experiments, in which the consensus of a

  dozen individuals had people doubting their own eyes, believ-

  ing that two lines of clearly diff er ent length were just as long,

  seem to offer the perfect demonstration of the power of major-

  ity opinion to quash prior beliefs, however strong they are. But

  Asch himself never framed his experiments in this way, stressing

  w h o k n o w s b e s t ? 75

  instead the power of the individual to resist group pressure.

  After all, the participants followed the crowd in only about one-

  third of the trials.36 Even then, it was the social pressure to con-

  form that drove this be hav ior, not the informational pull of the

  majority opinion.37 After the experiment was finished, many par-

  ticipants admitted to having yielded even though they knew the

  group to be wrong.38

  An even better proof rests with another version of the experi-

  ment in which participants were told that they had arrived late

  and, as a result, would have to write down their answers on a

  sheet of paper. Instead of having to voice their opinions in front

  of the group, which unanimously agreed on an obviously wrong

  line, they could answer privately. Conformity plummeted. In

  only a small minority of trials did participants choose to follow

  the consensus.39 Across all these experiments, a very small num-

  ber of people genuinely believed the group to be right, and they

  relied on a variety of strategi
es to make sense of the group’s weird

  answer— that the lines created a visual illusion, or that it was their

  width rather than their length that was the object of the task.40

  What about, then, Milgram’s conformity experiments, in

  which a few confederates looking up at a random building were

  seemingly able to get every passerby to do the same? A more re-

  cent take on this study shows that, unlike Asch’s participants,

  the passersby were not conforming because of social pressure.

  Psychologist Andrew Gallup and his colleagues replicated Mil-

  gram’s experiment, asking a varying number of confederates on

  busy sidewalks to look up in the same direction.41 This time, how-

  ever, the confederates were looking up at a camera and, using

  motion- tracking software, the researchers could describe in de-

  tail the be hav ior of the people around them. As in Milgram’s

  experiment, some people started looking up as well, but they

  were more likely to do so when they were behind, rather than

  76 ch ap t er 5

  in front of, the confederates. This rules out social pressure as a

  significant force, for then people would have started looking up

  when they could be seen and judged by the confederates. But

  Gallup also found that people reacted quite rationally on the

  whole, using sound cues to decide when to follow majority opin-

  ion. The passersby were more likely to look up when more con-

  federates were doing so, and when the crowd was thinner (and

  so a greater proportion of people were looking up). The research-

  ers also found that many people didn’t look up, showing the

  reaction not to be reflex- like at all. Instead, whether people

  looked up was likely determined by other factors, such as

  whether or not they were in a hurry.

  Competent Competence Detection

  On the whole, our mechanisms of open vigilance provide us with

  a good estimate of who knows best. Preschoolers already use an

  impressive number of cues to decide whether others might know

  more than they do. They consider who has the most reliable in-

  formational access. They use past per for mance to establish who

  is more competent in a given domain.42 They have a good idea

  of the bound aries of a specialist’s domain of expertise. They are

  more likely to follow the majority opinion when the majority is

  more numerous and closer to a consensus, when its members are

  competent, and when they have formed their opinions in de pen-

  dently of each other.43

  Preschoolers are open to changing their minds when they

  think others know better, yet they remain vigilant. Far from

  blindly following prestigious but incompetent individuals or the

  majority opinion, they weigh cues to competence and confor-

  mity against their prior beliefs, so that even in the face of experts

  or a strong consensus, they do not automatically change their

  w h o k n o w s b e s t ? 77

  minds— and neither do adults, contra a naive interpretation of

  the Asch conformity experiments.

  This doesn’t mean we cannot be tricked by people who seem

  competent or who are able to fake a consensus. Fool’s errands

  are an innocuous example. When an apprentice in a workshop

  is sent to look for elbow grease, all the cues tell him that he should

  comply. The workers asking him are competent, they all agree,

  and, if prompted, they provide in de pen dent reasons for the im-

  portance of the task at hand.44 To avoid looking like a fool, the

  apprentice has to ignore these cues and consider that all of these

  people might not have his best interests at heart.

  6

  WHO TO TRUST?

  in order to properly evaluate what we are being told, we

  need to figure out who knows best, but this isn’t enough. The

  most competent expert is useless if they decide to lie. A consen-

  sual group is of equally little use if its members conspire to de-

  ceive us.

  A tremendous amount of research has been dedicated to the

  question of “deception detection,” in other words, how good we

  are at spotting liars. Are we any good at it? What cues do we rely

  on? How reliable are these cues? The practical stakes seem huge.

  From human resource personnel to detectives, from betrayed

  spouses to victims of telephone scams, who wouldn’t like a fail-

  safe way to spot a liar?

  Subtle nonverbal cues are often thought to be reliable tells.

  Someone who fidgets, who looks shifty, or who is unwil ing to

  make eye contact is less likely to inspire confidence.1 For Freud,

  “No mortal can keep a secret. If his lips are silent, he chatters with

  his finger- tips; betrayal oozes out of him at every pore.”2

  Indeed, many people are quite confident in their ability to spot

  a liar. This is one of the reasons why, in many cultures, verbal

  testimony is preferred to written testimony in court: judges

  think they can tell if someone is lying by watching them speak

  78

  w h o t o t r us t ? 79

  in person.3 To this day, many detectives are taught to rely on

  “such visual cues as gaze aversion, nonfrontal posture, slouch-

  ing, and grooming gestures.”4

  The TV show Lie to Me rests on this premise. The show’s hero,

  Cal Lightman, is inspired by Paul Ekman, a psychologist famous

  for his studies of emotional expressions. Like Ekman, Lightman

  travels to faraway places in order to demonstrate that people

  make the same face to express, say, fear, everywhere in the world.

  Again like Ekman, Lightman uses his deep knowledge of emo-

  tional expressions to catch liars, in par tic u lar by relying on the

  observation of microexpressions.5

  Microexpressions are facial expressions that last for the blink

  of an eye, less than a fifth of a second. These rapid expressions

  are supposed to betray the conflicting emotions of those who try

  to lie, or to hide their feelings more generally. People who want

  to conceal their guilt, their sadness, or their joy might let slip a

  tiny muscle movement reflecting the emotion they attempt to

  hide rather than the one they want to display.

  Even though they are essential y invisible to the untrained eye,

  microexpressions would be perceptible by those who receive the

  proper instructions, such as the short class offered by Ekman to

  a variety of law enforcement agencies (and which you can easily

  buy online). It seems we’ve fi nally found a solution to catching

  liars, and it’s only a few hours’ training away.

  Microexpressions, Schmicroexpressions

  Unfortunately, things are not quite so simple. Ekman’s ideas,

  and his results, have proven controversial. Ekman’s critics point

  out that his findings about the reliability of microexpressions as

  a tool for spotting liars have not been published in proper

  80 ch ap t er 6

  peer- reviewed journals: he has not shared his methods and data

  with the scientific community, so they cannot be in de pen dently

  evaluated.6 Moreover, experiments conducted by scientists out-

  side his group have
yielded rather negative results.

  Psychologists Stephen Porter and Leanne ten Brinke showed

  participants stimuli designed to elicit a variety of emotions, from

  disgust to happiness, while asking some of the participants to

  display a diff er ent emotion than the one that would normally be

  elicited by the stimuli.7 They asked coders to judge the partici-

  pants’ facial expressions frame by frame (104,550 of them!) in

  order to detect the slightest slipups. In nearly a third of the trials

  in which they had to fake an emotion, participants displayed

  some inconsistent feelings: someone shown a disgusting image

  might express fear or happiness, however briefly.

  If this appears to vindicate Ekman’s theories, it really doesn’t,

  for two reasons. First, these slipups lasted at least one second on

  average, several times longer than microexpressions are supposed

  to last, making them easily perceptible by untrained observers.

  Second, out of the fourteen genuine microexpressions, six were

  displayed when participants were not attempting to hide any-

  thing. This makes microexpressions quite useless as a tool for

  detecting deception attempts. Another study by ten Brinke and

  her colleagues, in which participants displayed either genuine or

  fake remorse, yielded similar results: microexpressions were rare,

  and the participants who were genuinely remorseful were just

  as likely to display them as those who were faking it.8

  The prob lem extends beyond microexpressions. In the origi-

  nal study by Porter and ten Brinke, a third of the deceitful

  participants— those who had been asked to fake a facial

  expression— briefly displayed an expression diff er ent from the

  one they were supposed to display. But that was also true for

  27 percent of those who were not asked to fake anything. After

  w h o t o t r us t ? 81

  all, we often feel conflicting emotions.9 As a result, such slipups

  are completely unreliable for detecting deception, failing to catch

  most deceivers, and catching many who have nothing to hide.

  Porter and ten Brinke’s results fit a pattern. Dozens of studies

  have looked in detail at people as they tell lies or true stories,

  scanning every detail of their be hav ior for cues to deception.

  Meta- analyses of these experiments yield pretty grim results: no

  cue is strong enough to reliably tell who lies and who doesn’t.10