About RPM
CHIL@Rice

A brief description of the ACT-R/PM project.


Project Goals

ACT-R, while being rich in mechanisms for cognition, has a history of being somewhat weak in the perceptual-motor domain--a situation common in computational models of cognition. The "Visual Interface" was the first serious attempt to remedy the situation for ACT-R and was successful in many ways. However, the Visual Interface raised probably about as many issues as it resolved. ACT-R/PM is an attempt to address many of those issues and provide a more comprehensive theory of cognition and perceptual-motor performance.

In particular, ACT-R/PM is focused on:

This makes ACT-R/PM particularly well-suited for modeling high-performance human-machine interaction.

Ancestry

ACT-R/PM is a synthesis of John Anderson's ACT-R theory, Mike Matessa's Visual Interface for ACT-R, and Dave Kieras and Dave Meyer's EPIC system. ACT-R/PM uses the ACT-R production system/Bayesian network as its core cognitive model, and a parallel set of perceptual-motor modules (much like EPIC). The specific modules are derived from the Visual Interface and similar modules in EPIC.

Current Status

ACT-R/PM is now reasonably mature. It is still reasearch software, but the code base is stable and ACT-R/PM has been used to successfully model a variety of tasks. The ACT-R/PM distribution includes the source code, a manual, and example models built in ACT-R/PM. ACT-R/PM, like the Visual Interface before it, is able to interact directly with computer interfaces constructed in Macintosh Common Lisp (and now Allegro Common Lisp as well). For more information or to be included on the notification list for new releases, please contact byrne@acm.org.

Car

Last modified 2002.05.30