[ACT-R-users] Goal Stack

[db30 at andrew.cmu.edu]

--On Monday, January 12, 2015 5:40 PM +0000 "Kelley, Troy D CIV (US)" 
<troy.d.kelley6.civ at mail.mil> wrote:

> All,
>
> Could someone out there remind me of the reasoning for getting rid of the
> goal stack?
>
> We have been using a goal stack with SS-RICS and have found it to be useful
> for handling task interruptions.  So the stack is flexible and
> interruptible.
>
> If I remember correctly there was some question as to how the goal stack can
> have perfect recall, so that if one goal is popped, the next goal is
> automatically retrieved with no decay.
>
> We have solved this by having activation associated with goals, so that
> goals decay while they are on the stack.  This means that the goal might not
> be perfectly recalled to the goal stack once the previous goal is retrieved.
> This also duplicates the strange behavior of  - "walking into a room and not
> knowing why you are there".  Because the goal of going to a room to find
> some object has decayed while you were going to the room.
>
> Any ideas on this?
>
>


There may be better references out there, but the first three talks from
the 1999 ACT-R Workshop address this issue:

<http://act-r.psy.cmu.edu/?post_type=workshops&p=7369>

Basically, an infinitely deep perfect memory (the goal stack) is
not plausible, but using the existing declarative memory activation
mechanisms to retrieve past goals instead works well.

Hope that helps,
Dan