THink: Investigating Cognition

Human cognition is frequently explored with a variety of tests that involve some
form of writing (e.g., solving mazes, drawing objects, etc.). With the ability to
easily capture digitized pen strokes (e.g., with a digitizing ballpoint pen or a stylus
and tablet), the data available from these tests has suddenly become hundreds of
times more precise, opening up new windows into investigating cognition, both
healthy and impaired. Our multi-site study, underway for the past several years,
has been collecting data at 7 sites around the US, providing a robust body of
information that has already revealed new and unexpected phenomena.

The THink project is looking for both MEng and UROP students to participate in
this work on a range of tasks. We need people to help conceive of and implement
new measurements that take advantage of the precision available in the data. We
need people with an AI background and good Java skills to help make our
software much smarter in interpreting what people draw, i.e., figuring out what
each pen stroke is intended to be. This is a challenging problem involving state of
the art sketch interpretation and learning, using multiple representations, and a
variety of other very interesting issues. And we need people with a background
and interest in machine learning to help analyze the large body of data we have
collected.

If this sounds interesting, please get in touch with Professor Randall Davis
(davis@csail.mit.edu).