by Hugo Mercier
end up saying nothing, depriving us of potentially valuable in-
formation. After all, a tip from your foodie friend might still be
useful, even if she hasn’t taken every potential factor into account
in formulating her advice.
What we need is a way of managing expectations, a way for
senders to tell receivers how much weight to put on the messages.
This is the function of commitment signals.34 We can indicate how
much we commit to every thing we say (or write). When we com-
mit, we are essentially tel ing receivers that we are quite sure the
message will be valuable to them. Audiences should be more
likely to accept the message, but they should also react more
strongly if it turns out we haven’t been very diligent after all.
Commitment signals abound in human communication. Some
commitment markers are explicit indications of confidence: “I’m
90 ch ap t er 6
sure,” “I guess,” “I think,” and so forth. Epistemic modals—
might, may, could— also modulate commitment. Increased
confidence (and thus commitment) is also signaled implicitly
through nonlinguistic signals such as a greater variation in
pitch.35 Even the sources we provide for our beliefs have impli-
cations for our degree of commitment. Saying “I’ve seen that
Paula is pregnant” commits us more to the truth of Paula
being pregnant than does “ People say Paula is pregnant.” Even
young children are able to take these signals into account: for
example, two- year- olds are more likely to believe confident
speakers.36
Adjusting how much we take into account what people say as
a function of their level of commitment is sensible, but only if
two conditions are met: not every body’s commitments should
be treated equally, and we should adjust our future trust in light
of past commitment violations.
Elephants Do Not Forget (When People Have Been
Overconfident)
To properly take commitment into account, we must keep track
of how much our interlocutors value our continued cooperation,
so we know how much weight to put on their commitment. The
more we think they want to keep cooperating with us, the more
we can trust their commitment. But we must also keep track of
who is committed to what so that we can adjust our trust accord-
ingly. If commitment signals could be used wantonly, without
fear of repercussions, they would not be stable. In an attempt to
influence others, every body would constantly commit as much
as pos si ble, making the signals worthless.
In a series of experiments, psychologist Elizabeth Tenney and
her colleagues asked participants to deliver a verdict in a mock
w h o t o t r us t ? 91
trial on the basis of two testimonies, delivered by two witnesses,
one of which was more confident than the other.37 Because par-
ticipants had no other way of distinguishing between the testi-
monies, they put more faith in the more confident witness. Later
on, both witnesses were revealed to have been wrong, and par-
ticipants now found the witness who had been less confident to
be more credible: she had been as wrong as the other witness,
but less committed.
Colleagues and I have replicated and extended these find-
ings.38 Participants were exposed to two advisers who pro-
vided the same advice but with diff er ent degrees of confidence.
After it was revealed that both advisers had been wrong,
participants were more willing to trust, even in a completely
unrelated domain, the adviser who had been less confidently
wrong.
Saying that overconfidence doesn’t pay might seem odd.
Aren’t there many successful politicians or businesspeople who
are full of bluster? It is pos si ble that overconfidence might pay in
some settings, as when we lack good feedback on the speaker’s
actual per for mance. However, it is impor tant to keep in mind
that these settings are the exception rather than the rule. In small
groups, when people can monitor each other’s words and deeds
quite easily, overconfidence is a poor strategy. Small- scale tradi-
tional socie ties, which are relatively close to the environment
our ancestors lived in, offer a great example. When an individual
becomes a leader in such a society, it isn’t thanks to empty blus-
ter but because they have superior practical skil s, give better
advice, or know how to resolve conflicts.39 We also see this in
our everyday lives. Some of us might have been taken in by an
ebullient friend who swears their ideas are the best. But this is
short- lived. If their ideas don’t pan out, we learn to adjust our
expectations.
92 ch ap t er 6
We are more influenced— every thing else being equal—by
more committed speakers, those who express themselves more
confidently. But those more committed speakers lose more if it
turns out they were wrong. These costs in terms of lost reputa-
tion and ability to influence others are what keep commitment
signals stable.
Deciding Who to Trust
Deciding who to trust has little to do with looking for signs of
ner vous ness or attempting to spot elusive microexpressions. It
is not even primarily about catching liars. Instead, it is about rec-
ognizing who is diligent in communicating with us: who makes
the effort to provide information that is useful to us, rather than
only to them. And diligence is all about incentives: we can trust
speakers to be diligent when their incentives align with ours.
Incentives between sender and receiver are sometimes natu-
rally aligned— when people are in the same boat. However,
even minor misalignment in incentives can create communica-
tion breakdowns, so that naturally aligned incentives are rarely
sufficient on their own. To remedy this prob lem, we create our
own alignment in incentives by keeping track of who said what,
and by lowering our trust in those who provided unhelpful in-
formation. In turn, this monitoring motivates speakers to be dili-
gent in providing us information, creating a social alignment of
incentives.
Because we are able to track others’ commitments, and adjust
our trust accordingly, most of human communication isn’t cheap
talk but costly signals, as we pay a cost if our messages turn out
to have been unreliable. Arguably, it is this dynamic of commit-
ment that has allowed human communication to reach its
unpre ce dented scope and power. However, this ability to keep
w h o t o t r us t ? 93
track of who said what, as well as to figure out whether a speak-
er’s incentives are more or less aligned with ours, depends on
having access to a wealth of information. Through the majority
of our evolution, we would have known for most of our lives most
of the people we interacted with. As a result, we would have had
plenty of information to recognize aligned or misaligned incen-
tives; to spot overconfident, unreliable, deceitful individuals;
and to adjust accordingly how we value their commitment.
Ironically, even though we are now inundated with more infor-
mation than ever, we know very little about many of the people
whose actions affect us most. What do we know of the people who
make sure the products we buy are safe, who operate on us, who
fly our planes? We barely know the politicians who govern us,
besides what we can glean from scripted speeches and carefully
curated insights into their personal lives. How are we supposed,
then, to decide who to trust?
One of the organ izing princi ples of open vigilance mecha-
nisms is that, in the absence of positive cues, we reject com-
municated information: by default, people are conservative
rather than gullible. The same applies to trust. If we don’t know
anything about someone— indeed, if we don’t even know who
they are—we won’t trust them. A first step to establish trust is
thus to be recognized, as an individual or an entity. This is why
name recognition matters in politics, and branding matters in
marketing.40
Obviously, a name isn’t enough. To convince others of the reli-
ability of our messages, we have to do more. As we’ve seen in
the preceding chapters, good arguments, having access to rele-
vant information, or having performed well in the past can
make a speaker more convincing. However, trustworthiness is
often a bottleneck. Many arguments fall flat if we don’t accept
their premises on trust (e.g., that such and such study showing
94 ch ap t er 6
no link between autism and vaccines exists and is reliable). Even
the best- informed, most competent speaker shouldn’t be listened
to if we don’t think they have our interests at heart. Research on
advertising and po liti cal be hav ior lends some support to the im-
portance of trustworthiness. As we wil see in chapter 9, celebri-
ties help sell products mostly when they are perceived as expert
in the relevant domain, but trustworthiness matters even more.41
To the extent that the perceived personal traits of politicians
weigh on voting decisions, a study suggests the most impor tant
trait is “how much the candidate really cares about people like
you”—in other words, whether you think their incentives are
aligned with yours.42
The importance of trustworthiness is also highlighted by the
damage done when it breaks down. As a rule, the positive effect
that association with a (competent, trustworthy) celebrity might
have for a product is swamped by the fallout if negative informa-
tion about the celebrity emerges.43 For example, following the
revelations about Tiger Woods’s philandering, three brands that
had recruited him— Pepsi, Electronic Arts, and Nike— lost
nearly $6 bil ion in market value.44 Similarly, politicians are barely
rewarded for keeping their electoral promises (which they do
most of the time, at least in democracies), but they suffer high
electoral costs when they are convicted of corruption (as they
should).45
The relative paucity of information available to help us decide
who to trust means that, if anything, we do not trust enough— a
theme I will return to in chapter 15.
7
WHAT TO FEEL?
in the early 1960s, Tanganyika was in a state of flux. It had
declared its in de pen dence from British rule in 1961 but remained
part of the Commonwealth until 1962, when it shed that last link
with the United Kingdom. A year later, Tanganyika would unite
with neighboring Zanzibar to become the modern state of
Tanzania.
But in 1961, politics was not the only area of turmoil in Tan-
ganyika. In the Buboka district, on the western shore of Lake
Victoria, children were behaving weirdly. It all started on Janu-
ary 30, when three teenage girls, pupils at the same boarding
school, were beset with sudden, uncontrol able bouts of laugh-
ing and crying that lasted several hours.1 A year later, nearly one
hundred pupils were affected. The school was forced to close
down. The pupils went back home, spreading further their un-
usual and upsetting be hav iors. Over the coming months, the
contagion would affect hundreds of young people throughout
the district.
Outbreaks of bizarre be hav ior are hardly novel, and they are
far from over.2 In 2011, dozens of teenage girls from Le Roy, a
small town in upstate New York, were affected for months on end
with symptoms similar to those that had stricken the pupils in
Tanganyika fifty years earlier.3
95
96 ch ap t er 7
When describing these events, it is difficult to avoid analogies
with epidemiology: outbreak, spread, affected, contagion. The
two doctors who reported the events happening in Tanganyika
described them as “an epidemic of laughter.”4
The same analogies are often used to describe how people act
in crowds. In the late nineteenth century, contagion became the
dominant explanation for crowd be hav ior. Gustave Le Bon wrote
that “in crowds, ideas, sentiments, emotions, beliefs possess a
power of contagion as intense as that of microbes.”5 His colleague
Gabriel Tarde noted that “urban crowds are those in which con-
tagion is fastest, most intense, and most power ful.”6 The Italian
Scipio Sighele suggested that “moral contagion is as certain as
some physical diseases.”7
It is hardly surprising that the analogy between the spread of
emotions and pathogens flourished at that time. The second half
of the nineteenth century was the golden age of the germ the-
ory of disease, when it allowed John Snow to curb cholera epi-
demics; Louis Pasteur to develop inoculation against rabies; and
Robert Koch to identify the agents that cause anthrax, cholera,
and tuberculosis.8
The contagion analogy, as the germ theory of disease, would
only grow more successful. It has been used to describe panics,
such as the (supposed) reactions to Orson Welles’s “War of the
Worlds” broadcast, when thousands of listeners, thinking the
Martians had landed, are supposed to have fled in panic: the “first
viral- media event.”9 Armies are depicted as routing in “a conta-
gion of bewilderment and fear and ignorance.”10 Nowadays, talk
of contagion and viruses has become ubiquitous when describ-
ing the effects of social media— take the phrase viral marketing,
for instance— and not only in the popu lar press. In 2014, two
articles were published in the prestigious Proceedings of the
w h at t o f e e l? 97
National Acad emy of Sciences that attempted to detect and ma-
nipulate “emotional contagion [in] social networks.”11
There are indeed tantalizing parallels between the way dis-
eases, on the one hand, and emotions or be hav iors, on the
other, spread. People do not voluntarily emit pathogens. Simi-
larly, the expression of emotions or such bizarre be hav iors as
uncontrol able laughing or unstoppa
ble dancing are not under
voluntary control. People do not choose to be affected by patho-
gens. Similarly, people do not make a conscious decision to
start laughing or crying when they see others do the same—
indeed, in many cases they actively resist the urge. Some patho-
gens at least are extremely difficult to fight off. Likewise with
emotions and be hav iors in crowds: “Few can resist [their] con-
tagion,”12 suggested Nobel Prize– winning author Elias Canetti.
Fi nally, pathogens can have dire consequences. Similarly, behav-
ioral or emotional contagion can (or so it is said) “make of a
man a hero or an assassin”;13 it can lead individuals to “sacrifice
[their] personal interest for that of the collective.”14
If nineteenth- century psychologists relied chiefly on crude
observations of crowd be hav ior, their successors have performed
impressive experiments showing how fast and automatic the re-
action to emotional signals can be. Recording the movements
of their participants’ facial muscles, psychologists John Lanzetta
and Basil Englis showed that seeing someone smile or frown im-
mediately leads to the same muscles being activated in the ob-
server’s face.15 Later, psychologist Ulf Dimberg and his colleagues
showed that this automatic imitation happened even when the
emotional expressions were presented so quickly that they could
barely be consciously registered at all.16 This almost instan-
taneous activation of facial muscles is believed to be a sign of
contagion: people automatically take on the facial expressions
98 ch ap t er 7
of those they observe, which leads them to feel the same emo-
tion. Psychologist Guil aume Dezecache and his colleagues have
even shown this mimicry can extend to third parties: not only
do observers take on the emotional expressions of those they
observe, but those who observe the observers also adopt the
expression.17
With such results, it is hardly surprising that one of the most
influential books in the field of emotional communication, by
psychologists Elaine Hatfield, John Cacioppo, and Richard Rap-
son, should be titled Emotional Contagion, and bear on the
power of “rudimentary or primitive emotional contagion— that
which is relatively automatic, unintentional, uncontrol able, and
largely inaccessible to conversant awareness.”18
