[ACT-R-users] help with ACT-R 3

[Dan Bothell]

--On Sunday, June 29, 2003 11:49 AM -0400 Kevin Tor <gimpytor at yahoo.com> wrote:


>
> Hello,
>
> I recently inherited an ACT-R 3 model and I am working on cleaning up the
> code around it but, at some point, I would like to upgrade it to ACT-R 5.  I
> have two questions:
>
> 1.  Is there any way that I can still get info on ACT-R 3?
>

I don't know of anywhere that you can get ACT-R 3 materials, sorry.  It's been 
a while, but as I recall ACT-R 3 was very close to ACT-R 4 (anyone who 
remembers otherwise please correct me if I'm wrong there).  That would mean 
that the book "The Atomic Components of Thought" 
<http://act.psy.cmu.edu/book/> would be the best reference for the theory.


> 2.  Is there an easy way without rewriting the model from scratch to move it
> to ACT-R 5 without losing anything?
>

It would depend on the model, but generally speaking you will have to rewrite 
it if you want an actual ACT-R 5 model.  Because ACT-R 5 is mostly backward 
compatible with ACT-R 4 I suspect that it would also run an ACT-R 3 model as 
well, but just because the model runs using ACT-R 5 doesn't make it an "ACT-R 
5 model" .  One of the biggest differences between an ACT-R 4 (or 3) model and 
an ACT-R 5 model is how chunks are retrieved from declarative memory and 
depending on how much of that the model does will mostly determine the extent 
of the rewrite necessary. The difference is that prior to ACT-R 5 the 
retrievals occurred on the left hand side (LHS) of a production and there 
could be any number of them, but in ACT-R 5 a single retrieval is requested on 
the right hand side of one production and then harvested on the LHS of a later 
production.  Converting an older model to that style of operation is going to 
require rewriting probably most of the productions.

Hope that helps,
Dan