Page 320 - Deep Learning

The Formation of Belief 303

done by the hypothesis together with a supporting chorus of ill-distinguished
25
background beliefs.” The consequence is also the same: “When an observa-
tion shows that a system of beliefs must be overhauled, it leaves us to choose
which of those interlocking beliefs to revise.” But Quine follows the train of
26
thought a step further than Duhem: “Just about any hypothesis, after all, can
be held unrefuted no matter what, by making enough adjustments in other
27
beliefs.” Quine emphasized that facts and observations tend to impinge on a
belief system along its periphery, not at its center: “The totality of our so-called
knowledge or beliefs … is a man-made fabric which impinges on experience
only along the edges.” 28
Approaching the problem of resistance from the point of view of empiri-
cal social science rather than philosophy, Milton rokeach proposed that belief
29
systems exhibit a center­periphery structure. A belief system is normally orga-
nized around a core of fundamental and general beliefs. The latter tend to be
few in number, and they are often held implicitly. The belief that individuals
differ in their personality traits is an example. At a slightly lower level of gen-
erality, a person might adopt beliefs about particular personality traits; some
people are more outgoing than others, some people are risk takers and so on.
These more specific beliefs are consistent with the core belief but contain addi-
tional detail. Belts of increasingly specific beliefs are arranged in concentric
rings at successively increasing distances from the core. in each successive
ring, the beliefs become not only more specific, but also more numerous and
less important. At the periphery of the belief system there are highly particu-
lar beliefs such as the claim that risk takers are drawn to extreme sports and
risk takers are gamblers. The center-periphery structure can be seen either as
a hierarchy with the core beliefs at the apex and the peripheral beliefs at the
bottom – the sideways view, as it were – or as a set of concentric rings with
the core beliefs in the center and the peripheral beliefs in the outer circle – the
top-down view; see Figure 9.2.
rokeach hypothesized that resistance to change increases from periphery
30
to center. Peripheral beliefs change relatively easily, but core beliefs remain
stable. Like Quine, he emphasized that contradictory information will typi-
cally impact a belief system at its periphery, where the beliefs are most specific.
For example, a medical study is more likely to show that, for example, vitamin
C is not good for you (after all) than that vitamins (in general) are not good
for you. The amount of contradictory information required to overthrow an
established belief is therefore a function of the position of that belief within the
belief system. Change will happen first at the periphery, overthrowing specific
beliefs of narrow scope and little importance. Deeper contradictions might