Brown bag du 22/05/2017 Ivan Kroupin (Harvard University)



Problems of Induction in Abstract Thought: The Case Study of
Relational Match to Sample

It is uncontroversial that humans, especially adults, have a unique
capacity for flexible, abstract cognition. Much of this takes the form of
relational reasoning - reasoning with and comparison between the abstract
relations between sets of objects.
Relational reasoning underlies both analogy and metaphor and, by
extension, much of our scientific and artistic accomplishment as a
species. To date, the literature on relational reasoning has identified
two factors which account for adult humans' unique relational reasoning
prowess relative to young children and non-human animals:
1) Relational representations which are both far more numerous and
potentially of a different, more abstract structure and 2) Executive
function (e.g. working memory) capacity required to flexibly manipulate
these representations.
In this talk I will argue, first a priori and subsequently with empirical
evidence, that this cannot be the whole story. Specifically, I will first
outline how successful relational reasoning in a given context requires
not only the necessary representations and executive capacities, but also
the identification of which relations are meaningful in the given context.
Given that there is an infinity of relations in any context, this problem
of relational induction not a trivial challenge - even if the appropriate
representations and executive functions are in place.
I will then bring empirical data to bear on the problem, showing that
persistent failure on what is considered the 'gold standard' relational
reasoning task (Relational Match to Sample) can result from a failure of
relational induction and not representational or executive limitations -
at least in the case of four year old children and likely certain
non-human animals.
In conclusion, I will present the case that the mechanisms by which we
solve problems of relational induction - how we come to identify certain
relations as meaningful in specific situations - are a significant part of
why human adults' relational reasoning (and human-unique cognition
generally) is uniquely powerful and productive.
I will close by offering some preliminary speculation on the nature of
these mechanisms and discuss future directions.