[ACT-R-users] New ACT-R software available

[db30 at andrew.cmu.edu]

There is a new version of the ACT-R software available on the
ACT-R website:

<http://act-r.psy.cmu.edu/act-r-7-6/>

This is a working prototype of the system described at the 2017
ACT-R Workshop, and the slides of that talk are also available
from the ACT-R website (this system is covered in the second half
of the slides):

<http://act-r.psy.cmu.edu/wordpress/wp-content/uploads/2017/12/bothell-ACT-R-updates.pdf>


This is not a replacement for the current stable release which is
still version 7.5-<2244:2017-07-11>.  Although this is still a work
in progress, it is stable enough to run all of the tutorial tasks,
and we will be using it at CMU in the cognitive modeling class
this semester.  It is being made available incase others would
like to use it for teaching as well.

Why would one want to use an incomplete prototype for teaching,
you might ask.  The answer is because it's possible to work through
the entire tutorial without ever interacting directly with a Lisp
prompt.  All of the experiments and other miscellaneous code for
the tutorial units is available in both Lisp and Python, and the
tutorial texts have been updated to provide instructions for both
versions.  The model files however are still Lisp code.  The model
definitions have been separated from the task implementations, and
a single model file will work with either implementation of the task.

The interface which allows the tasks to be written in Python is a
general protocol for external access, and it should be possible to
write a client for it in any language in which one can open a socket
and parse JSON data.  The docs for that low-level interface are
included in the docs directory in the remote.pdf file, and in
addition to the Python client used for the tutorial tasks there is
another, simpler, Python client example available in the file:
extras/creating-modules/external/goal_complete.py (only available
with the source code distribution not the standalones).  That
example implements a replacement for the ACT-R goal module written
in Python which is complete enough to run most of the tutorial tasks
as-is using that goal module instead of the built-in Lisp version.

Now, all that said, there are of course reasons one might not want
to use this (other than the obvious of not wanting to use prototype
grade software).  One is that the only valid documentation on using
this version of the software currently is in the tutorial units --
none of the old documentation files in the docs directory have been
updated appropriately yet.  Also, as you may have guessed from the
change in the major version number (7.5 -> 7.6) there are changes
to the software which may cause problems for existing ACT-R code.
In particular, the whole device concept for perceptual/motor
interfacing had to be reworked.  Because of that, basically any
task which uses a device is going to need some changes before it'll
work in this version.

If you do try it out, please let me know of any problems or issues
that you encounter.

Dan