the rule "If people get their faces scarified then they have a
right to eat buffalo," subjects spontaneously check for buffalo-
eaters with intact faces (rather than scarified individuals who
do not eat buffalo). Inferences in such situations follow a spe-
cific "check for cheaters" rule rather than a general logic.
Indeed, subjects are confused when asked to check an equiva-
lent rule that does not pertain to social exchange, such as "If
people get their faces scarified then they have visited Peking."
Psychologists observed these same experimental results in
American college students and Shiwiar hunter-gatherers in
the Amazon. That social exchange is a special inference sys- [125]
tem is confirmed by the fact that it can be disrupted by brain
pathology while other reasoning functions are preserved.^27
Evaluation of trust. That humans depend on cooperation
createsstrategicproblems, where the value (the expected bene-
fit) of a particular move depends on whether someone else
makes a particular move (not necessarily the same one). Ide-
ally, one could choose to cooperate only with people who are
forced to cooperate. Whenever you pay for something you
can of course draw a gun and threaten the shopkeeper to be
sure that you get the correct change. This is not really possi-
ble in most circumstances, but we have capacities that com-
pensate for that. One is that we can decide to cooperate with
people on the basis of particular signalsfrom which (we think)
we can infer that they are cooperators.^28
Now this part of the computation is crucial, but we are not
really aware of it. This situation is in fact general. Sociologists
Diego Gambetta and Paul Bacharach have studied extensively
the signals by which people evaluate other people's trustworthi-
ness in everyday situations. They show that in many contexts
(e.g., Would you let the person who arrived just behind you
into the building, even though she has not dialed the appropri-
ate code or rung the bell?) people are able to evaluate whether
certain signals are reliable or not. This requires that they com-
pute both the significance of the signal and the probability that
it is faked. All this is done automatically and quickly: not
because the inferences are simple but because we have special-
ized systems that carry out this computational work.29
Coalitional dynamics. This is another common feature of
human interaction. People will spontaneously form groups
THE KIND OF MIND ITTAKES