4.1 Cheater detection: a module for monitoring social exchangeFamiliarity of the subject matter may have some eVect on performance,
but it is swamped by the divergence on deontic (normative) and indicative
(descriptive) versions of the selection task. Cosmides and Tooby (1992)
have provided a decisive demonstration of this by testing subjects on tasks
involving conditionals such as ‘If you eat duiker meat, then you have
found an ostrich eggshell’ and ‘If a man eats cassava root, then he must
have a tattoo on his face’. Results showed that subjects performed better
on deontic tasks than on indicative tasks concerning the same subject
matter, whether that subject matter was familiar or not. Cosmides and
Tooby’s theory is that we have a specialised cognitive mechanism – a
cheater-detection module– for reasoning about the sorts of deontic tasks on
which subjects perform well. They consider these deontic tasks to engage a
system of reasoning which is dedicated to normative conditionals con-
cerning social contracts of the form ‘If you have received the beneWt, then
you must pay the cost’, or of the form ‘If you have paid the cost, then you
should receive the beneWt’.
On a modularist and nativist approach of the type we advocate, there is
a particularly strong case for postulating something like a cheater-detec-
tion system, on grounds of evolutionary selective pressures. For it is clear
that human evolution has been heavily dependent upon novel forms of
social co-operation. Yet along with all the advantages of co-operation
comes an obvious danger – without special cognitive adaptations for
monitoring social exchange, co-operative humans would oVer to the free-
riding cheat, who takes the beneWts without paying the costs, an oppor-
tunity for exploitation. So, on game-theoretic grounds, a capacity to spot
the cheater would be badly needed and likely to evolve. To do the job
required, this would have to be a central module, capable of taking input
from a variety of peripheral systems (such as hearing what people say and
seeing what they are doing, as well as recollection of previous agreements).
Hence the proposal of a special reasoning module for ‘cheater-detection’.
Actually, this line of thought underestimates the evolutionary signiW-
cance of such a special reasoning module. For it cannot be right to suppose
thatWrstthere was contract-based social co-operation,thensome individ-
uals start to cheat, andthenthe others develop a capacity to detect cheating
and take appropriate steps to exclude cheaters from the beneWts. What is
wrong with this picture is that innovative forms of social co-operation
could not develop at all without a cognitive adaptation for social ex-
change. Before any worries about how to police a social contract arise, one
needs a social contract to police. But that involves an understanding, on
the part of contracting individuals, of what the contract implies. Those
120 Reasoning and irrationality