PM-Question

[Dan Bothell]

--On Wednesday, July 24, 2002 1:38 PM +0200 Wolfgang Schoppek 
<Wolfgang.Schoppek at uni-bayreuth.de> wrote:

> This is a more technical question concerning ACT-R/PM (2.1b6).
> There are 8 visual-locations in my display. Some of the objects in
> these locations change with every new items, some of them stay the
> same. The model looks for locations with new objects with the
> following productions:
>
> (P find-next-feature
>     =goal>
>        ISA wm-free
>        state find
> ==>
>     +visual-location>
>        ISA visual-location
>        attended nil
>     =goal>
>        state looking
> )
>
> (P repeat-find-next-feature
>     =goal>
>        ISA wm-free
>        state looking
>     =visual-location>
>        ISA visual-location
>        attended t
> ==>
>     +visual-location>
>        ISA visual-location
>        attended nil
> )
>
> "Now" these are the visual locations:
>
> P11    1.564
>     isa VISUAL-LOCATION
>     attended t
>
> P21    1.403
>     isa VISUAL-LOCATION
>     attended t
>
> P31    1.428
>     isa VISUAL-LOCATION
>     attended t
>
> P41    1.334
>     isa VISUAL-LOCATION
>     attended t
>
> P12    1.969
>     isa VISUAL-LOCATION
>     attended t
>
> P22    1.607
>     isa VISUAL-LOCATION
>     attended t
>
> P32    1.356
>     isa VISUAL-LOCATION
>     attended t
>
> P42    1.328
>     isa VISUAL-LOCATION
>     attended nil
>
> The problem is that even though there is one location which has not
> been attended (p42), PM keeps pushing p12 into the visual-location
> buffer. This effect occurs with and without partial matching.
> A second question: When there is a new item on the display with, say,
> 4 changes, only one of the visual-locations changes to "attended new".
> The other keep their "attended t". Nevertheless, PM scans at least 2
> or 3 of the new locations. New visual-location chunks never show up in
> the declarative memory viewer.
>
> Any hints?
>

The declarative memory doesn't automatically update with PM changes. You 
should use the pm-print-icon command to get the actual state of the visicon, 
which is probably quite different from what you're seeing in the declarative 
memory viewer.

Dan