Hi Everyone,
We are excited to announce that the next CRAM session will take place this THURSDAY, February 13th, at
12-1PM in room 461 (2001 McGill College). We will have the pleasure of hearing from Dr. Brendan Johns,
our department's latest faculty hire. See below for the title and abstract of the talk.
As usual, snacks will be provided. All are welcome.
Best Regards,
The CRAM Team
----------------------
Title: The Continued Importance of Theory: Lessons from Big Data Approaches to Cognition
Abstract: The cognitive sciences have begun to collect, collate, and use large natural language corpora to
understand human lexical behavior. That effort is facilitated by impressive new machine learning
techniques and technologies. The exercise has advanced the field and produced insights that would have
otherwise remained hidden. However, it is important to consider the tradition and practice of big data within
the history of cognitive science and to consider the potential costs associated with an over enthusiastic
and unreflective embrace of the new way. By tradition, cognitive science is an empirical science grounded
in the experimental hypothetico-deductive method. In short, theories produce a priori predictions and
those predictions are tested in controlled targeted experiments. To the extent that the data match the
theoretical predictions, the theories are sound. When data mismatch the theory, the data prevail. In
contrast, the standard practice in big data is abductive: explanations are provided for found data. As any
logician would tell you, deduction and abduction are not equivalent. In light of the current scenario, that
difference forces a meaningful but sometimes overlooked distinction in the purpose as well as explanatory
role of classical versus big data methods for understanding human behavior. I will articulate the distinction
and present examples using computational models of language processing.