Much of what we do at Adaptive Cognitive Systems is focused in one way or another on Cognitive Modeling, the science of using computational simulation as a tool for understanding, predicting, and emulating human behavior.
Cognitive Modeling is distinguished from regular programming, or
Machine Learning, for that matter, in its emphasis on modeling not just
the observed behavior of an individual, but the processes
involved. The core scientific methods have moved from using
verbal reports as data to dependence on neuroscience, eye-tracking, and
fMRI as ways of measuring and identifying human behavior and cognitive
processes, but the basic methodology is the same.
This basic methodology involves identifying a target data set for modeling containing measures of human behavior in some task environment, and then developing a program that produces a close approximation to that behavior, relying on processes that are known to be processes humans are capable of engaging in.
A successful model produces a good approximation, not just for the
data set used in building the model, but for either a portion of the
data that was not seen during model construction (a "split-half"
methodology), or for a completely new data set (extrapolation).
The measure of success, then, will be some measure of fit to the new