[ACT-R-users] puzzled by act-r 5.0.6 behavior

Dan Bothell db30 at andrew.cmu.edu
Sat Nov 30 01:54:30 EST 2002


--On Friday, November 29, 2002 4:27 PM -0500 Bruno Emond 
<bruno.emond at nrc-cnrc.gc.ca> wrote:

> I am puzzled by a retrieval behavior in ACT-R 5.
> It seems that there is a range of activation values for which
> neither a chunk nor the failure chunked are retrieved.
>

What's happening is that the next production is firing before the retrieval 
completes, thus canceling the previous attempt.  In the trace the latency 
displayed shows when that retrieval would complete if it wasn't canceled.  So, 
here for instance

> /*
> Then there is the problematic section. Between an activation of 2.926 and
> 2.000 nil is returned, not chunk x2 or failure. Even below and activation of
> 2.000 nil is returned, contrary to the previous example.
> */
>
>   Sources of activation are: nil
>      Computing a base level of  2.926 from 1 references from creation time
> 0.000      Computing a base level of  2.926 from 1 references from creation
> time  0.000   CHUNK X2 Activation  2.926 Latency  0.054
>   Time  0.500: Ret Fired
>   Time  0.500: Ret Selected
> ==> NIL
>   Sources of activation are: nil
>      Computing a base level of  2.841 from 1 references from creation time
> 0.000      Computing a base level of  2.841 from 1 references from creation
> time  0.000   CHUNK X2 Activation  2.841 Latency  0.058
>   Time  0.550: Ret Fired
>   Time  0.550: Ret Selected

chunk X2 will be retrieved at time .554, but the production is selected and 
fired again at time .550 so the retrieval doesn't get a chance to complete.

Hope that clears things up,
Dan





More information about the ACT-R-users mailing list