Page 373 - Deep Learning

356 Conversion

to ontological knowledge,* but I suggest that the power of miscategorization to
distort information and the consequent need for recategorization are possibili-
ties associated with any type of knowledge.
These differences notwithstanding, the ontological shift and resubsump-
tion theories share a fundamental assumption: In both, the cure for the initial
miscategorization of an experience or a discourse is to recategorize it under
some alternative higher-order knowledge structure. Conversion is a process
of reorganizing knowledge, not a process of falsification. Both theories imply
that the alternative higher-order knowledge representation must already exist
for the change to take place and therefore postulate parallel development of
such higher-order representations prior to the change. The point cannot be
emphasized enough: According to the Resubsumption Theory, a new theory is
not a product of theory change, but one of its prerequisites.
Unlike theories that tackle the effects of resistance head on, the
Resubsumption Theory resolves the question of where the contender theory
comes from by designating the presence in memory of an alternative theory
as one of the triggering factors for theory change. No special assumptions are
needed to explain the origin of the contender theory: It was formed by the
same processes as any belief system, because it was not formed as an alterna-
tive but as part of the person’s ongoing efforts to make sense of experience and
discourse. The possibility of an alternative is, in part, a fact about the world –
domains of experience sometimes share enough structure to be subsumed
under one and the same theory – and, in part, a fact about the mind: Due to
the layered nature of mental representations, a belief system sometimes turns
out to be abstract enough to subsume another domain than the one for which
it was created. It is the positive step of creating an alternative theory, in con-
junction with a background theory that claims that the two theories are incom-
patible, that leads to the rejection of the resident theory, not a confrontation
with evidence. A theory might replace another without either theory suffering
from any contradiction with observations or other types of evidence.
Unlike most theories of conversion, the Resubsumption Theory explains
why conversions are rare without invoking the processes of resistance described
in Chapter 9. Although the mechanisms of the Resubsumption Theory afford a
route to belief revision that circumvents resistance to contradictory informa-
tion, it does not imply that conversion is common or easy to induce. Conversion
occurs only under the simultaneous occurrence of multiple conditions. Even

* Chi and Brem (2009) de-emphasize their prior focus on ontological knowledge, but it is not
clear how this affects the relation between category shift and resubsumption.