by Hugo Mercier
For instance, if you, like most people, got stuck on the ten-cents
answer to the Bat and Ball prob lem, and someone had told you
that the correct answer was five cents, your initial reaction would
have been to reject their statement. In this case, your System 2
would have had to do some work to make you accept a sound
belief, which is much more typical than cases in which System 2
does extra work to make us reject an unfounded belief.
There is no experimental evidence suggesting a systematic as-
sociation between being less analytical y inclined— less likely
to use one’s System 2— and being more likely to accept empiri-
cally dubious beliefs. Instead, we observe a complex interaction
between people’s inclinations to use diff er ent cognitive mecha-
nisms and the type of empirically dubious beliefs they accept.
Beliefs that resonate with people’s background views should be
more successful among those who rely less on System 2, whether
e v o l v in g o p e n - mind e d ne s s 45
or not these beliefs are correct. But an overreliance on System 2
can also lead to the ac cep tance of questionable beliefs that stem
from seemingly strong, but in fact flawed, arguments.
This is what we observe: the association between analytic
thinking and the ac cep tance of empirically dubious beliefs is
anything but straightforward. Analytic thinking is related to
atheism but only in some countries.51 In Japan, being more
analytically inclined is correlated with a greater ac cep tance of
paranormal beliefs.52 Where brainwashing techniques failed to
convert any POWs to the virtues of communism, the sophisti-
cated arguments of Marx and Engels convinced a fair number
of Western thinkers. Indeed, intellectuals are usually the first to
accept new and apparently implausible ideas. Many of these
ideas have been proven right (from plate tectonics to quantum
physics), but a large number have been misguided (from cold
fusion to the humoral theory of disease).
Even when relative lack of sophistication seems to coincide
with gullibility, there is no evidence to suggest the former is caus-
ing the latter. On some mea sures, young children can be said
to be more gullible than their older peers or than adults.53 For
instance, it is difficult for three- year- olds to understand that
someone is lying to them and to stop trusting them.54 (In other
re spects, obviously, three- year- olds are incredibly pigheaded,
as any parent who has tried to get their toddler to eat broccoli
or go to bed early knows). But this apparent (and partial) gull-
ibility isn’t caused by a lack of cognitive maturation. Instead, it
reflects the realities of toddlers’ environment: compared with
adults, small children know very little, and they can usually trust
what the adults around them say.55 In the environment in which
we evolved, young children were nearly always in the vicinity of
their mothers, who have limited incentive to deceive them, and
who would have prevented most abuse. This strong assumption
46 ch ap t er 3
of trustworthiness adopted by young children is, in some ways,
similar to that found in bees, which have even fewer reasons to
mistrust other bees than young children have to mistrust their
caregivers. In neither case does lack of sophistication play any
explanatory role in why some agents trust or do not trust others.
The logic of evolution makes it essentially impossible for gull-
ibility to be a stable trait. Gullible individuals would be taken
advantage of until they stop paying attention to messages. In-
stead, humans have to be vigilant. An arms race view of the
evolution of vigilance is intuitively appealing, with senders evolv-
ing to manipulate receivers, and receivers evolving to ward off
these attempts. Even though this arms race view parallels nicely
the popu lar association between lack of sophistication and gull-
ibility, it is mistaken. Instead, openness and vigilance evolved
hand in hand as human communication became increasingly
broad and power ful. We can now explore in more detail the cog-
nitive mechanisms that allow us to be both open and vigilant
toward communication: How do we decide what to believe, who
knows best, who to trust, and what to feel.
4
WHAT TO BELIEVE?
imagine you are a foodie. You love all sorts of diff er ent
cuisines. There’s one exception, though: Swiss cuisine. Based on
a number of experiences, you have come to think it is mediocre
at best. Then your friend Jacques tell you that a new Swiss res-
taurant has opened in the neighborhood, and that it is really
good. What do you do?
Even such a mundane piece of communication illustrates the
variety of cues that you ought to consider when evaluating any
message. Has Jacques been to the restaurant, or has he just heard
about it? Does he particularly like Swiss cuisine, or is he knowl-
edgeable about food in general? Does he have shares in this new
venture? The next two chapters are devoted to identifying and
understanding the cues that relate to the source of the message.
Here I focus on the content of the message.
Imagine that Jacques is as knowledgeable about eating out as
you are and has no reason to oversell this restaurant. How do you
integrate his point of view— that the Swiss restaurant is
great— with your skepticism toward Swiss cuisine? Evaluating
messages in light of our preexisting beliefs is the task of the most
basic open vigilance mechanism: plausibility checking.
On the one hand, it is obvious enough that we should use our
preexisting views and knowledge when evaluating what we’re
47
48 ch ap t er 4
told. If someone tells you the moon is made of cheese, some
skepticism is called for. If you have consistently had positive in-
teractions with Juanita over the years, and someone tel s you
she has been a complete jerk to them, you should treat that piece
of information with caution.
On the other hand, doesn’t relying on our preexisting beliefs
open the door to bias? If we reject every thing that conflicts with
our preexisting views, don’t we become hopelessly stubborn and
prejudiced?
How to Deal with Contrary Opinions
Experimental evidence suggests the risk of irrational stub-
bornness is real. In some circumstances, people seem to be-
come even more entrenched in their views when presented
with contrary evidence—to use the earlier example, it is as if
you would become even more sure that Swiss cuisine sucks
after being told that a Swiss restaurant was great. Psycholo-
gists call this phenomenon the backfire effect. It has been re-
peatedly observed; for instance, in an experiment that took
place in the years following the second Iraq War. U.S. presi-
dent George W. Bush and his government had invoked as a
reason for invading Iraq the supposed development of weap-
ons of mass destruction by
Iraqi leader Saddam Hussein. Even
though no such weapons were ever found, the belief they existed
persisted for years, especially among conservatives, who had
been more likely to support Bush and the Iraq War. In this
context, po liti cal scientists Brendan Nyhan and Jason Reifler
presented American conservatives with authoritative infor-
mation about the absence of weapons of mass destruction in
Iraq.1 Instead of changing their minds in light of this new infor-
mation, even a little bit, the participants became more convinced
w h at t o b e l ie v e ? 49
that there had been weapons of mass destruction. A few years
later, the same researchers would observe a similar effect
among staunch opponents of vaccination: presenting anti-
vaxxers with information on the safety and usefulness of the
flu vaccine lowered even further their intention of getting the
flu shot.2
Surely, though, the backfire effect has to be the exception
rather than the rule. Imagine you’re asked to guess the length of
the Nile. You think it is about seven thousand kilo meters long.
Someone says it is closer to five thousand kilo meters. If the back-
fire effect were the rule, after several more iterations of the argu-
ment, you would be saying the Nile is long enough to circle the
earth several times over. Fortunately, that doesn’t happen. In this
kind of situation— you think the Nile is seven thousand kilo-
meters long, someone else thinks it is five thousand— people
move about a third of the way toward the other opinion and very
rarely away from it.3
Even on sensitive issues, such as politics or health, backfire
effects are very rare. Nyhan and Reifler had shown that conser-
vatives told about the absence of weapons of mass destruction
in Iraq had become even more convinced of the weapons’ exis-
tence. Po liti cal scientists Thomas Wood and Ethan Porter re-
cently attempted to replicate this finding. They succeeded but
found this was the only instance of a backfire effect out of thirty
persuasion attempts. In the twenty- nine other cases, in which
participants were provided with a factual statement relating to
U.S. politics (for example, that gun vio lence has declined, or that
there are fewer abortions than ever), their opinions moved in line
with the new, reliable information. This was true even when the
information went against their preexisting opinion and their po-
liti cal stance.4 As a rule, when people are presented with mes-
sages from credible sources that challenge their views, they move
50 ch ap t er 4
some of the way toward incorporating this new information into
their worldview.5
In the examples we’ve seen so far, there was a direct clash
between people’s beliefs (that there were weapons of mass
destruction in Iraq, say) and what they were told (that there
were no such weapons). The case of the Swiss restaurant is a little
subtler. You don’t have an opinion about the specific restaurant
Jacques recommended, only a prejudice against Swiss cuisine in
general. In this case, the best thing to do is somewhat counter-
intuitive. On the one hand, you are justified in doubting Jacques’s
opinion and thinking this new Swiss restaurant is likely to be bad.
But you shouldn’t then become even more sure that Swiss cui-
sine in general is poor— that would be a backfire effect. Instead,
your beliefs about Swiss cuisine in general should become some-
what less negative, so that if enough (competent and trustwor-
thy) people tell you that Swiss restaurants are great, you end up
changing your mind.6
Beyond Plausibility Checking: Argumentation
Plausibility checking is an ever- present filter, weighing on
whether messages are accepted or rejected. On the whole, this
filtering role is mostly negative. If plausibility checking lets in
only messages that fit with our prior beliefs, not much change
of mind is going to occur— since we already essentially agree
with the message. This is why you often need to recognize the
qualities of a source of information— that it is reliable and of
goodwill—to change your mind. There is, however, an exception,
a case in which plausibility checking on its own, with no infor-
mation whatsoever about the source, gives us a reason to accept
a novel piece of information: when the new information in-
creases the coherence of our beliefs.7
w h at t o b e l ie v e ? 51
Insight prob lems are good examples of how a new piece of
information can be accepted purely based on its content. Take
the following prob lem:
Ciara and Saoirse were born on the same day of the same
month of the same year to the same mother and the same father
yet they are not twins.
How is this pos si ble?
If you don’t already know the answer, give it a minute or two.
Now imagine someone saying “ They’re part of triplets.” Even
if you have no trust whatsoever in the individual tel ing you this,
and even though this is a new piece of information, you will ac-
cept the answer. It just makes sense: by resolving the inconsis-
tency between two girls being born at the same time of the same
mother and their not being twins, it makes your beliefs more
coherent.
In some cases, simply being told something is not enough to
change our minds, even though accepting the information would
make our beliefs more coherent. Take the following prob lem:
Paul is looking at Linda.
Linda is looking at John.
Paul is married but John is not.
Is a person who is married looking at a person who is not
married?
Yes / No / Cannot be determined.
Think about it for as long as you like (it is one of my favorite
logical puzzles, which my colleagues and I have used in many
experiments).8
Now that you’ve settled on an answer, imagine that your friend
Chetana tel s you, “The correct answer is Yes.” Unless you happen
52 ch ap t er 4
to already think Yes is the correct answer, you will likely believe
Chetana has gotten the prob lem completely wrong. You prob-
ably reached the conclusion that the correct answer is Cannot be
determined— indeed, you are likely quite sure this is the correct
answer.9
Yet Chetana would be right, and you would be better off ac-
cepting Yes as a correct answer. Why? Because Linda must be
either married or not married. If she is married, then it is true
that someone who is married (Linda) is looking at someone who
is not married (John). But if she is not married, it is also true that
someone who is married (Paul) is looking at someone who is not
married (Linda). Because it is always true that someone who is
married is looking at someone who is not married, the correct
answer is Yes.
Once you accept the Yes answer, you are better off. Yet, among
the people who initially provide the wrong answer (so, the vast
ma
jority of people), essentially no one accepts the answer Yes if
they are just told so without the accompanying argument.10 They
need the reason to help them connect the dots.
Arguments aren’t only useful for logical prob lems; they are
also omnipresent in everyday life. Going to visit a client with a
colleague, you plan on taking the metro line 6 to reach your des-
tination. She suggests taking the bus instead. You point out that
the metro would be faster, but she reminds you the metro con-
ductors are on strike, convincing you to take the bus. If you had
not accepted her argument, you would have gone to the metro,
found it closed, and wasted valuable time.
The cognitive mechanism people rely on to evaluate argu-
ments can be called reasoning. Reasoning gives you intuitions
about the quality of arguments. When you hear the argument
for the Yes answer, or for why you should take the bus, reason-
ing tel s you that these are good arguments that warrant chang-
w h at t o b e l ie v e ? 53
ing your mind. The same mechanism is used when we attempt
to convince others, as we consider potential arguments with
which to reach that end.11
Reasoning works in a way that is very similar to plausibility
checking. Plausibility checking uses our preexisting beliefs to
evaluate what we’re told. Reasoning uses our preexisting infer-
ential mechanisms instead. The argument that you shouldn’t take
the metro because the conductors are on strike works because
you naturally draw inferences between “The conductors are on
strike” and “The metro will be closed” to “We can’t take the
metro.”12 If you had thought of the strike yourself, you would
have drawn the same inference and accepted the same conclu-
sion: your colleague was just helping you connect the dots.
In the metro case, the dots are very easy to connect, and you
might have done so without help. In other instances, however,
the task is much harder, as in the prob lem with Linda, Paul, and
John. A new mathematical proof connects dots in a way that is
entirely novel and very hard to reach, yet the people who under-
stand the proof only need their preexisting intuitions about the
validity of each step to evaluate it.
This view of reasoning helps explain the debates surrounding
the Socratic method. In Plato’s Meno, Socrates walks a young