Page 334 - Deep Learning

The Formation of Belief 317

triggering condition for change as the appearance of a scientific problem, and
“problems arise … where our ideas about the world are at variance either with
nature, or with one another: … Conceptual problems in science emerge from
the comparison, not of ‘propositions’ with ‘observations’, but of ‘ideas’ with
59
‘experience.’ ” A conceptual innovation “calls … for a collective dissatisfac-
tion with the existing conceptual repertory.” Toulmin does not explain how
60
an “idea” being “at variance” with “experience” differs from Kuhn’s concept of
an anomaly.
The sheer reasonableness of the idea that theory change is driven by con-
flicting evidence is the strongest argument in its favor. Why would scientists
search for a better theory unless they recognize flaws in their current one?
But the assumption of cognitive conflict as a driving force brings with it the
need to explain how resistance is overcome. Why are the anomalous, con-
tradictory or falsifying data not swept under the cognitive rug via periph-
eral changes? What enables contradictory information to travel up a belief
hierarchy and affect core scientific principles? Although cognitive analyses
of science have generated many different proposals regarding the processes
of theory change, there is no new and clearly articulated answer to this ques-
tion. For example, there have been several attempts to propose a Darwinian
variation-selection theory of scientific progress. According to Toulmin, the
key features of the growth of scientific knowledge are the generation of a pop-
ulation of conceptual variants and their selection for transmission to the next
generation of scholars. Presumably, contradictory data and arguments serve
the selective function in this application of variation-selection. But why are
the data and arguments accepted rather than resisted? A variation-selection
account presupposes that negative information about conceptual variants is
heeded rather than resisted, but it does not by itself explain how or when,
under which circumstances, this happens.
other theoretical proposals fare no better on this point. Philip Kitcher’s
1993 book Advancement of Science argues that science moves forward by
61
increasing its explanatory power. in Kitcher’s view, a scientific practice is
organized around a repertoire of explanation schemas. An explanation schema
is a template for how to structure the explanation of a particular type of phe-
nomenon or event. For example, Darwinian variation-selection explanation
should identify the ancestor species, describe the heritable variations, explain
the relevant selective pressure and calculate its cumulative effect over many
generations. To make scientific progress is to improve the repertoire of expla-
nation schemas. Kitcher writes, “Four distinct kinds of processes are at work
[in] explanatory progress.” They are “the introduction of correct schemata”;