by Hugo Mercier
attention to such information. This is true even when the threats
are reported in rumors. Nearly a year before Pearl Harbor, the
U.S. ambassador to Japan heard that plans for an attack were
being hatched, but he dismissed the rumors as unreliable, with
158 ch ap t er 10
devastating consequences.34 As a result of these costs asymme-
tries, information about threats is often deemed relevant even
if it is not practically relevant. Rumors about the toll of natu ral
disasters, lurking sexual predators, or conspiracies in our
midst are bizarre forms of mind candy: guilty pleasures that
might not be good for us, yet we can’t help but enjoy.35
Conspiracy theories are a salient form of threat. Given the im-
portance of co ali tions during our evolution, it is plausible that we
could have evolved to be particularly attuned to the risk raised by
an alliance forming against us.36 Even if we do not have anything
like a dedicated “conspiracy detector,” conspiracy theories com-
bine ele ments that make them relevant: they are about a co ali tion
(jackpot) of power ful people (double jackpot) who represent a
significant threat to us ( triple jackpot).
A study that looked at more than a hundred thousand rumors
on Twitter found that the most successful false rumors (com-
pared with true rumors) were those that elicited disgust and
surprise.37 The logic of cost asymmetry applies to disgust: as a
rule, it is better to find too many things disgusting (thereby avoid-
ing potential pathogens) than too few. As for surprise, it is sim-
ply a general mea sure of relevance: every thing else being equal,
more surprising information is more relevant information. Our
five- and seven- year- old sons encapsulated this wisdom while
they were discussing which pieces of rubbish found on the beach
they liked most: “the things that are disgusting, and that we don’t
know, that’s what’s in ter est ing.”
By this logic, times of crisis are prone to rumormongering not
because crises make people more gullible but because they make
people curious about topics they had no interest in previously,
from the amount of rainfal to a rabies epidemic on the other side
of the state. These new sources of relevance are often not matched
by an increase in practical relevance, so that false rumors can
t i t il l at in g r um o r s 159
spread with relative ease. By contrast, if the rumors have serious
practical consequences, people do their best to check them,
whether or not there is a crisis situation.
Rewarding Rumor Relays
We tend to reward people who provide us with relevant infor-
mation: we like them more and think of them as more compe-
tent and more helpful.38 In order to win as many of these rewards
as pos si ble, we should be able to ascertain the social relevance
of a piece of information— how valuable it is for other people—
so that we know what to transmit.
Sometimes this means recognizing that a piece of information
would be relevant for a specific individual: if you know a friend
of yours is a big Lego fan, it is good to realize information about
a Lego exhibit would be relevant for her. To compute this nar-
row social relevance, we rely on our knowledge of the preferences
and beliefs of specific individuals.
In other cases, we’re more interested in figuring out whether
a piece of information would be relevant for many people. To
compute this broad social relevance, we use our own mind as a
guide: information that happens to be relevant for us is deemed
relevant for others ( whether or not it has any practical conse-
quences). This isn’t a trivial pro cess. When nonhuman animals
encounter information that triggers many inferences— the trail
of a prey, signs of a predator— they don’t think to themselves,
“Damn, that’s in ter est ing!” They just perform the required op-
erations and engage in the appropriate be hav ior. Humans, by
contrast, are able to represent the relevance of the stimuli they
encounter. When we hear something shocking on the news, we
don’t simply adjust our beliefs; we also take note of this piece of
information’s broad relevance, so we can share it later. The same
160 ch ap t er 10
goes for jokes, stories, tips— and rumors. When a rumor taps
into the actual domains of many cognitive mechanisms, such as
those related to threats or conspiracies, we realize that the rumor
is likely to have broad social relevance.
Information that has broad social relevance has a special prop-
erty: it becomes more valuable because of its own relevance.
As a rule, you score social points when you give people infor-
mation they find relevant. The friend you told about the Lego
exhibit will note that you are thoughtful and well informed,
because this information was useful to her. But what if this in-
formation were useful not just to her but to other people she
knows? Then she would, in turn, be able to score social points
by spreading this information. When we give people informa-
tion that has broad social relevance, we score double points:
points because they find the information relevant, and more
points because they can use the information to score points in
turn, and they are thankful for it.
News that we get through mass media often has broad social
relevance (that’s why the media are talking about it), but because
the media have a wide reach, it can be difficult to make much
social use of them. By contrast, rumors are the perfect material
to appear in ter est ing: because they typically spread to only one
or a handful of individuals at each step, they provide us with
plenty of opportunities to score social points— not only because
our interlocutors find the rumor relevant but also because it wil
help them score points in turn.
Minimal Plausibility
Rumors tap into the actual domain of many cognitive mecha-
nisms, giving them the potential to be highly relevant. But to
be genuinely relevant, they also have to be plausible. Some
t i t il l at in g r um o r s 161
communicated content, from jokes to fairy tales, can spread
successfully without being considered true, or even plausible.
This is not the case for rumors. “ Children are being sexually
abused in a restaurant!” is neither funny nor entertaining. If it is
not at least somewhat plausible, it is completely uninteresting.
If we like rumors in part because they allow us to score social
points by spreading them, we should be careful that we don’t
spread implausible rumors. Or, more precisely, we should be
careful that we don’t spread rumors that others find implau-
sible. If we do, we not only fail to gain any benefit but also
might suffer some costs as a result: as a rule, people will not
trust us as much if we provide them with information deemed
implausible.
As I have argued in chapters 3 to 7, when it comes to evaluat-
; ing what others tell us, open vigilance mechanisms are mostly
on the lookout for cues that the message should be accepted.
Absent such cues, the default is rejection. Given the risks of
communication— accepting the wrong piece of information
could be life- threatening— this is a safe and sensible way to
operate.
By contrast, when guessing whether someone else is going to
accept or reject a piece of information, the potential costs are
lower. In the vast majority of situations, the risk of saying some-
thing wrong is being thought a little less smart, a little less dili-
gent. Only when people suspect duplicity, and when the stakes
are high, could the costs be prohibitive— and this is not the case
with most false rumors. These social costs should not be ne-
glected, but they are much lower than the potential costs of
being misled by others. As a result, when attempting to estimate
how others evaluate our messages, we can use the opposite strategy
to that used when we evaluate theirs: look for cues the message
might be rejected, rather than cues it will be accepted. In the absence
162 ch ap t er 10
of any indication that our interlocutors are predisposed to reject
the message, our default is to think they will accept it.
The way people spread the rumeur d’Orléans fits well with this
indirect use of open vigilance mechanisms. People seemed to
have no qualms about spreading the rumor far and wide. Only
when the citizens of Orléans thought someone would be in a
good position to reject the rumor did they choose to not share
it with them. No one directly shared the rumor with the town’s
older, better integrated Jewish population, who might have
personally known the suspected shop keep ers, and thus be in a
position to refute the rumor. Likewise, the only people to call the
police were those who genuinely wondered whether the rumor
was true
Counterintuitively, open vigilance mechanisms are actually
doing their job when it comes to false rumors. First, they evalu-
ate the rumor to decide whether it should be accepted, and
their verdict is: not really. This leads to a merely reflective ac cep-
tance, precluding costly behavioral consequences. Second, the
same mechanisms are used in a roundabout manner, to gauge
the odds that others would reject the rumor, helping us avoid
the social costs of sharing the rumor with those who might
deem it ridicu lous, and thus judge us negatively.
Escape from Real ity
Rumors take diff er ent forms and circulate in diff er ent ecosys-
tems. At one extreme we find rumors that are practically rele-
vant for at least some of those involved— who will get fired or
promoted, sent to the front or repatriated. This practical rele-
vance motivates people to circulate the information, as they
gain social points for doing so. But it also makes them conscious
that the information they provide will be checked by knowledge-
t i t il l at in g r um o r s 163
able individuals and, if found wanting, damage the reputation
of those who shared it. Rumors that circulate in this way are over-
whelmingly likely to be accurate.
At the other extreme we find rumors that are of little prac-
tical consequence, but that people think have broad social
relevance. These mind-candy rumors titillate our interest in
information about famous people, threats, conspiracies, and so
forth. The fact that they are mind candy has several conse-
quences. It means not only that we are interested in them but
also that we expect others to be interested in them. We even
expect our interlocutors to be interested in them because their
interlocutors will be interested in them. By spreading a juicy
rumor, we give people an opportunity to spread it further, for
which we get some kudos.
Several factors make it difficult for these rumors to be cor-
rected. The lack of practical consequences means that the be-
liefs don’t properly interact with the actual world, traveling in a
world of expectations about what other people find in ter est ing
that remains one step removed from real ity.39 As rumors and real-
ity come into contact, those who hold false rumors are promptly
disabused. When Edgar Maddison Welch acted on the basis of
the rumors about Comet Ping Pong, he quickly suffered the con-
sequences. Had more people done so, the rumor would never
have had a chance to spread.
Moreover, not only the benefits but also the costs of sharing
these rumors are social. To avoid these costs, we withhold the
rumors from those most likely to prove skeptical, further lower-
ing the chances of getting negative feedback. One of the reasons
the rumeur d’Orléans proved so successful is that the inhabitants
studiously avoided disclosing the rumor to policemen or to any-
one who might know better, further reducing the chances that
the rumor interacts with real ity.
164 ch ap t er 10
Conspiracy theories go even further in hampering good feed-
back, as the people in the best position to know whether there
is a conspiracy are those accused of engaging in the conspiracy.
But, obviously, if the conspiracy were true, the conspirators
wouldn’t admit to it, making any denial suspicious. Again, the
rumeur d’Orléans provides an example of how easy it is for rumors
to turn into accusations of conspiracy, and even for the fight
against rumors to take a conspiracist turn. In its initial form, the
rumor wasn’t a standard conspiracy theory, because the pre-
sumed evildoers— the shop keep ers accused of kidnapping
girls— weren’t in a position of power. But as the rumor grew, it
started facing contradictions: If these crimes were widely known,
why weren’t the police doing anything? They must have been
paid off. The politicians who attempted to debunk the rumor
must also be in on it. It is only then that the rumor was ready to
col apse under its own weight: too many people knew local po-
licemen or politicians for the rumor to be credible any longer,
at least for the majority of the population.
What to Do?
If false rumors spread so well, it isn’t because people take
them too seriously but because they don’t take them seriously
enough.
To increase the ratio of true to false rumors, then, we should
try to close the gap between social and practical relevance. When
we find a rumor appealing, we should pause before spreading it
further—by gossiping to a friend or hitting retweet. What would
we do if we had to make a practical decision based on this rumor?
Would we engage in vigilante justice to stop the sexual abuse of
children? Thinking of a rumor in practical terms should moti-
vate us to check it further.
t i t il l at in g r um o r s 165
Some information, however, can hardly be made to have prac-
tical consequences. For instance, the exaggerated stories that
appear in the wake of a natu ral disast
er are hardly actionable. In
these cases, we can at least imagine sharing the information with
someone who would be in a good position to know, or who
would be affected by the information (which is often the same
thing). Many would-be truthers would balk at the idea of shar-
ing their suspicions with 9/11 first responders. Someone who
happily ventures the figure of $140 million as a settlement for
David Dao’s misfortune to shock his friends might fear looking
stupid in front of an experienced personal injury lawyer.
Imagining the personal costs of acting on the basis of a rumor
is a good first step in the fight against false rumors.40 At least,
we’re less likely to be part of the prob lem. To be part of the solu-
tion, we should try to inflict some costs on those who spread
false rumors—at the very least deny them any benefits. We
shouldn’t hesitate to raise doubts, to question the plausibility of
their stories or the reliability of their sources (the importance of
which I will explore in the next chapter). This typically entails
a social cost. After all, people like a juicy rumor, which they can
use to score social points. We don’t thank the skeptic who spoils
it for every one. To reduce this cost, and to increase the effective-
ness of our questioning, we should be as polite as pos si ble, re-
strain from imputing nefarious intentions to those who spread
questionable rumors, and be careful not to overstate our own
claims. Are we really sure the rumor is false?
We should think of these personal costs as a contribution to
the public good. We would all be a little bit better off if fewer false
rumors circulated. However, to reach such an equilibrium, many
people will have to deprive themselves of potential benefits, re-
fusing to spread juicy rumors and incurring the small cost of
being thought a skeptical kil joy.
11
FROM CIRCULAR REPORTING TO
SUPER NATURAL BELIEFS
One of the major factors that distinguishes accurate and
inaccurate rumors, and that I haven’t touched on yet, is the qual-
ity of sourcing. By sourcing I mean providing our interlocutors
with a description of how we got a piece of information.
In environments conducive to the spread of accurate rumors,
people say things like “Bill Smith from HQ said that John was
