Page 324 - Deep Learning

The Formation of Belief 307

might decide that a newspaper that reports uncomfortable facts deliberately
distorts them to propagandize a particular point of view. This move is read-
ily observable in the public discourse about political and social matters. The
wider the scope of such a belief, the more convenient it is: if the entire news-
paper staff, not merely a particular journalist, is believed to be systematically
biased, then one has given oneself permission to ignore every uncomfortable
fact reported by that newspaper. A closely related move is to declare the source
to be mistaken. one can admit the honesty of the source and yet claim that the
facts are not as stated: The journalist did not interview the right people; the
interviewees lied for their own reasons; and so on. similarly, a research study
that produced an unwelcome finding can be dismissed by deciding that it is
methodologically flawed. Adding beliefs about the source of the contradictory
information rather than about its content allows one to resolve the contradic-
tion without revising prior beliefs and without incorporating the new infor-
mation into the belief base.

Bolstering
one of Festinger’s more interesting proposals was the hypothesis that mental
states differ in degree of dissonance and that the dissonance is proportional to
the ratio between contradicting and supporting cognitive elements related to
38
the contested belief. in network terms, a belief is a node, supporting argu-
ments or facts are connected to the belief node via links with a positive value
and contradicting assertions and arguments are connected via links with a
negative value. Festinger hypothesized that the degree of dissonance associ-
ated with a belief is proportional to the ratio of the number of negative to
the number of positive links. if a belief is contradicted by two arguments and
supported by three, the dissonance is of the same strength as if it were contra-
dicted by four arguments and supported by six.
These hypotheses imply that it is possible to reduce the dissonance asso-
ciated with a conflict by adding supporting evidence, without either revising
the content of the contradicted belief or processing the arguments against it.
if the strength of the dissonance associated with a belief is 8/4, and we find
8 new arguments in favor, the dissonance is reduced to 8/12. on the basis of
this principle, Festinger predicted that a car buyer will tend to spend more
time reading consumer reports that praise the car after the purchase than
before so that he can deal with any dissonance caused by negative comments
or consumer reports about the car. Fifty years of experimental research by
39
social psychologists has supported this and other counterintuitive predictions
derived from the bolstering idea. The effects of this mechanism are visible in
40