Learning new production rules in ACT-R

Update: the learning mechanism described here is part of ACT-R 5.0, the version of the architecture. I recommend using that version. However, the code given below can still be used as an extension to ACT-R 4.0.

This page contains files associated with a new proposal for learning new production rules in ACT-R. Note however, that the version offered here is still an experimental version that may contain bugs, or can be changed withour prior notice.

However, we very much welcome that people try this version out, and send us their comments, experiences and bug reports.

The code has mainly been tested using MCL, but also appears to work with ACL.

The following files are available (or download them all as Macintosh archive):

• Documentation (word file)
• The lisp-code, to be loaded after ACT-R 4.0
• Example code (all explained in the documentation)
  • simple-add-demo
  • simple-add
  • add-counting
  • paired-associative

A more elaborate example, learning the past tense, can be found in the published models page at CMU, and is described in the paper found in our prepublications page.