Page 339 - Deep Learning

322 Conversion

contradictory information. But why is that information not swept under the
rug via peripheral dissonance reduction? Where does the alternative belief
come from? Under which circumstances is it adopted as true?

ontological Category shifts

The ontological shift Hypothesis developed by Michelene T. H. Chi and co-
workers provides a different view of cognitive change. They focus on the role
74
of ontological knowledge in conceptual change. The theory assumes that the
learner possesses a small set of high-level categories that specify the types of
entities the person believes exist: event, mental state, object, process, value and
so on. such ontological categories are defined in terms of the predicates (prop-
erties) that are meaningful to assert or deny about the category members.
75
For example, every object has weight but it does not make sense to say that
some process, such as rain, weighs 10 kilograms. The problem is not that the
statement is false because the true weight is, say, 11 kilograms instead, but that
the predicate “weighs X kilograms” cannot be meaningfully asserted or denied
about rain or other processes.
When the learner acquires information about some previously unfamil-
iar phenomenon, that phenomenon gets assigned to whatever ontological
category seems appropriate on the basis of easily accessible features. The
phenomenon inherits the main features of that category. in particular, the
ontological category controls which predicates can be applied to the phe-
nomenon, which in turn has consequences for how new information about
the phenomenon is understood. Presumably, assertions that employ predi-
cates that are not meaningful for the assigned ontological category will not
make sense to the recipient. indeed, the rain today weighed 10 kilograms can
only be understood as referring to the amount of rainwater collected in a
particular container, or else in some metaphorical way. if the learner assigns
a phenomenon to the wrong category initially, he or she will tend to mis-
understand information about it. The cure for this is an ontological shift, a
re-assignment of the phenomenon to a different ontological category. if the
learner already is in possession of the correct ontological category, the rel-
evant processes are to detach the phenomenon from one category and attach
it to another. if the learner does not already possess the correct ontological
category, that category has to be learned before the shift can occur. Unless
the learner is prompted to create the missing category, the misconception
caused by its absence is likely to be robust.