[ACT-R-users] ACT-UP Beta Test

[David Reitter]

I'm soliciting beta-testers for ACT-UP, a toolbox implementation of ACT-R.

ACT-UP provides a programmer's interface to cognitive models.  ACT-UP models are very different from ACT-R models in that their task strategies are specified in Lisp instructions rather than production rules.  ACT-UP uses functions such as `retrieve-chunk' and `learn-chunk' to access declarative memory, implementing all of DM's functions (including blending and associative learning).  ACT-UP also provides reinforcement learning (utility learning) and compilation (akin production compilation) mechanisms.  Yet, ACT-UP is not a faithful reimplementation of ACT-R 6.  Its goal is to provide rapid prototyping, scalability (multi-agent modeling and large-scale, complex models), and ease of model re-use.

Detailed information and downloads are available at

http://act-up.psy.cmu.edu/

ACT-UP is well-documented and has been extensively tested with new implementations of the canonical models from the ACT-R tutorial (included).  We have an ACT-UP tutorial for you, as well as API documentation.  Slides from a talk and an ICCM paper are also available for download.

If you'd like to beta-test it, I ask that you implement a model of your own in ACT-UP and report back to us what works and what doesn't.  Moreover, I am interested in your professional opinion about it, and where we should take it.  Just e-mail me to join the test team. 


--
Dr. David Reitter
Postdoctoral Researcher
Department of Psychology
Carnegie Mellon University
http://www.david-reitter.com