which our belief-forming processes can be assessed, besides those oftruth
andspeed(Stich, 1990). We do not share this view, at least as an explication
of our ordinary notion of ‘rationality’ – except in so far as what counts as
enoughtrue beliefs may depend upon our background needs and purposes
as agents; but it would take us too far aWeld to argue for this here. In what
follows we shall assume that processes should be assessed in relation to
bothreliabilityandfecundity(truth and speed). And for the most part we
shall focus on just one belief-forming process in particular – namely
inferences assessing the truth of conditionals – as an example from which,
we hope, general morals can be drawn.
2 Some psychological evidence
There is a considerable body of evidence which shows that over a range of
reasoning tasks people perform poorly. Reasoning about probabilities
seems to be an area of particular weakness. Thegambler’s fallacyis a piece
of probabilistic irrationality to which people seem strongly prone. This
fallacy, of supposing that the probability of an independent event can
somehow be aVected by a series of previous occurrences (for example, the
coin having just come up heads three times in a row), is strangely seductive
even for those who ought to know better. Tests have also revealed the
popularity of theconjunction fallacy– judging the probability of a con-
junction, such as ‘Albert is a physicist and an atheist’, as being higher than
the probability of either of the two conjuncts (‘Albert is a physicist’,
‘Albert is an atheist’) in relation to the same background information
(Tversky and Kahneman, 1983). Other tests have been carried out to see
how belief is aVected by evidence which is subsequently discredited. If the
question is put explicitly, probably everyone will agree that one ought to
modify one’s beliefs once the grounds for holding them are shown to be
false – you ought not to go on believing something you no longer have
reason to believe. However, though people mightsaythey ought not to go
on believing, it has been found they have a tendency to do just that,
exhibitingbelief perseveranceeven after the bogus nature of some concoc-
ted ‘evidence’ has been carefully explained to them in a debrieWng session
(Nisbett and Ross, 1980).
It might be questioned whether results such as these could reveal any
general defect in human reasoning. For people can be taught reasoning
skills, can’t they? Once given adequate training, surely they will perform
much better on reasoning tests than untutored and unsuspecting subjects?
The answer is that they may well do so, but there is no guarantee they will
not revert to bad habits of reasoning as soon as they start to deal with cases
other than the examples on which they have been instructed. This is
108 Reasoning and irrationality