As well as controlling where and how we search for information, we can choose how
long we spend looking for new evidence related to a given issue. Another way that confirmation bias might manifest in search, therefore, is if people spend differing amounts
of time seeking out new information depending on how much support they presently have for their current belief. Confirmation bias would predict that people will spend more
time searching for information if the initial search is not favourable towards their current belief, than if they quickly obtain supportive information.
Though little research has looked at this directly, there is some more indirect evidence that people apply different standards of evaluation to information that confirms than
disconfirms their current beliefs (as I will discuss in more detail in the inference section.) This then suggests that people will spend more time searching for information if it sug-
gests an opinion-inconsistent conclusion than if it initially supports an opinion-consistent conclusion. Ditto and Lopez (1992) find that subjects spend more time questioning the
results of a test, and are more likely to re-take the test to gain more information, if they believe the test indicates a negative diagnosis. This also highlights the fact the bias in
search and bias in inference are more closely related than is sometimes acknowledged - since the inferences we draw affect our beliefs, which in turn influence our motivation
and assumptions in seeking out new information (as indicated by the arrow from the end back to the start of the process in figure 2.1, and as I will discuss in more detail in
2.2.3).
To see why this might result in a confirmation bias, imagine that of all scientific studies
conducted on gun control, there are in fact only three supporting the claim that gun control reduces homicides, ten that find no evidence of such a relationship, and five
that find evidence of the opposite relationship. If I already believe that gun control is effective, the evidence I end up being exposed to - and therefore whether I change my
mind or not - may end up being highly contingent on whatever I happen to read first. If I read a pro gun-control study first, I may be so convinced of the conclusion that I decide
to stop there - whereas if I read one of the studies suggesting the opposite relationship first, my confusion may lead me to continue reading more, meaning I ultimately sample
more of the available information and am better informed about the topic.
The evidence for this tendency to spend more time seeking out information if the initial
results are not confirmatory comes primarily from the literature on motivated reasoning (Kunda, 1990), which subtly differs from confirmation bias. The claim of motivated
reasoning is that people will reason in ways biased towards whatever they want to believe, (as opposed to confirmation bias which claims a bias towards whatever people currently
believe.) Is it reasonable to expect this tendency would generalise to confirmation bias? Assuming that people want to continue believing whatever it is they currently believe -
a fairly reasonable-seeming assumption - we might then expect findings from motivated reasoning to generalise to confirmation bias. However, it’s worth recognising that there
is a leap being made here, and the evidence does not directly support the claim for the case of confirmation bias.
2.2.1.5 ‘Myside bias’ in producing arguments
A final way in which confirmation bias has been suggested to occur in the search for in- formation is in theproduction of arguments: people supposedly find it easier to generate arguments for their preferred position than for the alternative. This seems appropriately categorised as ‘bias in search’ since we might think of producing arguments as searching
memory for reasons - an asymmetry in the arguments one produces therefore might be thought of as biased memory search.
Toplak and Stanovich (2003) asked undergraduate participants to generate arguments on both sides of three issues, and found that people consistently generated more ‘my-
side’ arguments than other side arguments. Perkins et al. (1986) found a similar effect. However, Wolfe and Britt point out that “with respect to written argumentation, it is
not at all clear that the tendency to generate more arguments for myside is an adequate definition of the myside bias, or even if it constitutes a bias of any kind.” (Wolfe and
Britt, 2008, p.3) Whether or not this can reasonably be considered a bias presumably depends on what people believe the purpose of generating arguments is: if trying to
put a case forwards for one side is even a plausible interpretation of the task, then an asymmetry in arguments presented may be a very reasonable response. To understand
bias in the generation of arguments, it’s helpful to distinguish different settings: gener- ating arguments in order to come to a conclusion about an important question is very
different from doing so in the context of defending a specific position, for example.
Wolfe and Britt (2008) define myside bias in written argumentation more strictly, as a
failure to make any reference to other side arguments or positions. In two experiments, they find that many participants only discuss one side of a given issue when asked to
put forward an argument, but that this arisesindependently of their personal opinions. If asked to write an essay in favour of a proposal, many participants fail to discuss any
arguments against the proposal, regardless of what their personal beliefs are on the issue. The authors suggest that the ‘myside bias’ is therefore not rooted in a ‘personal opinion
bias’ or confirmation bias - but rather stems from misperceptions about the nature of argumentation. Questionable beliefs about thinking (Baron, 1991, 1995) - such as the
impression one-sided arguments are stronger than two-sided arguments - may better account for what is going on here than a confirmation bias per se.