|
 |
|
Computational modeling is a tool for simulating
the cognitive processes underlying human behavior in a formal way.
It allows us to think concretely about these processes and forces
us to be explicit about the underlying mechanisms. We use artificial
neural network model developed in the connectionist, or parallel distributed
processing framework. This allows us to simulate human performance
in cognitive tasks using principles of processing that are similar
to those believed to apply in the brain. Thus, information is represented
as graded patterns of activity over populations of simple simple units,
processing takes place as the flow of activity from one set of units
to another, and learning occurs through the modification of the connection
strengths between these. From one perspective, such models are highly
simplified, capturing brain-style compu-
|
 |
tation, without necessarily committing to the
details of any particular neural system or sub-system. However, with
appropriate refinement, such models offer the opportunity to build
bridges between our understanding of the low-level properties of neural
systems, and their participation in higher level (system) behavior.
Along these lines, we have begun to refine our models of cognitive
performance, taking account of increasingly detailed information at
the neurobiological level.
Our modeling work explores the mechanisms by which the brain maintains
context and allocates attention in the face of different task demands.
A central hypothesis of this exploration is that context can be simulated
as an internal source of activation to a the system that is self-maintaining
and can serves to bias processing in other components of the system.
We have used this idea to simulate detailed aspects of human behavior
in a variety of cognitive tasks, including the Stroop task, the Eriksen
flankers task, the continuous performance test (CPT), and a variety
of others involving attentional control. We have also applied these
modeling principles to explore other cognitive tasks typically associated
with frontal function such as working memory, behavioral inhibition,
and their breakdown in schizophrenia. |
|