ACT-R/AC Serial Subtraction Model
Homepage

Research Team: Frank E. Ritter1, Karen S. Quigley3, Laura Cousino Klein2, Michael McNeese1,
Dirk Van Rooy1, Isaac Councill1, Marios N. Avraamides4, Michele McClellan Stine2,
and Isabella Rodriguez2

1 School of Information Science and Technology
Penn State, University Park, PA 16802

2 Department of Biobehavioral Health
Penn State, University  Park, PA 16802

3 Department of Psychiatry
University of Medicine and Dentistry of New Jersey, Newark, NJ 07103
and
VA Medical Center
New Jersey Health Care System
East Orange, NJ 07018

4 Department of Psychology
Penn State, University Park, PA 16802
 

Model and displays written by Isaac Councill
E-mail: igc2@psu.edu

 

Project Overview

A next step in improving the fit of cognitive models to human performance is to include the effects of behavioral moderators within cognitive architectures. These effects can be included by modifying the knowledge in the model, modifying architectural parameters, and by modifying the mechanisms within the architecture itself. We provide an example overlay to the ACT-R architecture that illustrates the first two of these approaches, including the effects of anxiety realized as worry and task-appraisal. This overlay is applied to an ACT-R model that performs serial subtraction. The resulting behavior matches the existing available data on human behavior on pre-task appraisal and serial subtraction and makes new predictions.

Software Downloads

Documentation
 

Quicktime Movies

Model running with a threatened pre-task appraisal, with worry turned on.
File size 5.4 MB
 

Model running with challenged pre-task appraisal.
File Size 2.1 MB
 

Screenshot

Model running with challenged pre-task appraisal and worry turned on.

The above image presents the graphical interface of the model. The two main windows are the Control Panel and the Model Behavior windows. The Control Panel window contains several options for selecting the model's conditions, run control, and some advanced output options. This window facilitates the setting of the model's moderators.

The Model Behavior window displays the current result and whether it is correct or erroneous, as well as the declarative memory chunks that are being used to solve the problem. Summary statistics (number of attempts, number of errors, and task latency) are also displayed in this window.





Applied Cognitive Science Lab
School of Information Science and Technology
Penn State University

Last modified January 20, 2002