What is Cognitive Modelling?

We are in a fun part of the history of cognitive science. We have a relatively short history, ~80 years, although some of the component disciplines are ancient. There is a lot to explore here. Given the interdisciplinary nature of the field, we’re just starting to integrate the disciplines from the outset of training. Only in the last two decades (give or take an exception) have cognitive models really had a seat at the table in the component disciplines of cognitive science. In truth, we still have growing pains. We’ve been capitalizing on the replication crisis in psychology to make it clear where cognitive science fits into the scientific process.

If you want one take on this integration, I recommend Olivia Guest and Andrea Martin’s Math Psych talk^[Math Psych is Australian for computational cognitive psychology.], which you can find here (~15 minutes). It also gives a glimpse at the sociology of cognitive modellers and experimental psychologists. In their talk, they paint two caricatures of the scientific process.

The first one reflects the inductive role of cognitive modelling.

  • We start by observing some data.
  • We construct a hypothesis based on that data.
  • We write some code.
  • We translate it into math.
  • We craft a theory to explain the math and assumptions.
  • We build a framework that removes some of the assumptions so that we can apply our theory to new situations and tasks.

The second caricature takes this process in reverse and reflects the traditional deductive experimental approach.

  • We start with a framework.
  • We apply the framework to a specific problem/task as a theory.
  • We specify the math and the assumptions of our theory.
  • We implement this specification in code.
  • We generate hypotheses from our implementation.
  • We conduct an experiment to test our hypothesis.

Now these are caricatures, which means they reflect life in exaggerated unrealistic ways. Olivia and Andrea argue that really we’re constantly moving in both directions. Science is as a path function linking theories to data. The red shaded steps in their diagram highlight the frequent path of the modeller. We often begin with some high level framework and empirical data and want to develop theories that explain the data and generalize to new tasks and situations.

Now this process might seem complicated and that’s okay. This is day one. We are going to go slower and we’re going to keep things light. This is an intro course.