[ACT-R-users] Baseline activation strengthening and decay

Wayne Gray grayw at rpi.edu
Thu Feb 16 10:16:47 EST 2006


Ben,

Don't forget that retrieval is very likely to take longer than 50 
msec. 50 ms is just the time it takes for the production that 
harvests the retrieval to fire. There would be a retrieval interval 
of anything from 100 to 1000 ms before a successful retrieval would 
place the item into the retrieval buffer. The retrieval latency is a 
function of baselevel activation as well as any temporary boost in 
activation the item may receive from the goal chunk.

1. production fires to initiate retrieval -- 50ms
2. memory system successfully retrieves the DME -- 100 to 1000 ms
3. production fires that harvests the item from the retrieval buffer -- 50 ms

In terms of what could be going on in your example, I think that some 
of Erik Altmann's work on task switching might be relevant as he has 
developed an idiom for rehearsing items to a certain level of 
activation.

In actr you can do a retrieval either with or without activation from 
the goal chunk. The activation from the goal chunk provides a 
temporary boost that enables items to be retrieved that do not have 
enough baselevel activation to be retrieved without the boost.

Hence, Erik has the system play a game with itself in which it tries 
to retrieve an item without the goal chunk  activation. If it cannot, 
then it attempts retrieval with goal chunk activation. A successful 
retrieval boosts activation. Then the system tries again to retrieve 
without that extra boost from the goal chunk --  when this retrieval 
is successful the system stops the rehearsal loop  -- in ACTR 4.0 
this all worked very nicely. Of course, if the system cannot retrieve 
the item even with the boost from the goal chunk, then it could 
always look at the screen again and recode it. I forget whether this 
was an issue for Erik's models.

I believe the above is a simplification as it has been some time 
since I looked at that code. But you can imagine that if a retrieval 
failure occurs that the system would simply try again -- as 
activation is noisy something that is below threshold on one 
retrieval attempt may be above retrieval threshold on another. The 
retrieval threshold is adjustable. I am not sure that there is a 
consensus on what this should be, but I believe that by default the 
system waits 1000 ms before deciding that the retrieval failed. (This 
is way too long and most modelers seem to lower it.)

Hope this helps.

Wayne




At 22:30 -0500 2006/02/15, ben.willems at faa.gov wrote:
>Please excuse my ignorance, but I am working on the other side of 
>the cognitive modelers.  That is, I run simulations using human 
>experts in Air Traffic Control and create records of human activity 
>quite similar to what you call a simulation trace.  Ignoring for a 
>moment that the visual system may be able to process several things 
>within the foveal area simultaneously, how would you interpret a 
>single 1500msec fixation on an object in terms of number of 
>retrievals?  Does that include a single retrieval or does it involve 
>cyclic retrievals with a time constant of 50msec for the retrieval 
>and another 50msec to push the chunk to the goal stack?  Or do you 
>assume that initially there is a retrieval followed by maintaining 
>activation at a faster cycle time?  Do you assume that activation 
>strengthening occurs independent of the perceptual or motor event 
>that triggers activation of the chunk?  E.g., seeing an aircraft 
>representation vs. listening to a reference to an aircraft or typing 
>in an identifier for that aircraft.
>
>Ben Willems
>Engineering Research Psychologist
>William J. Hughes Technical Center
>NAS Human Factors Group (ACB-220)
>Building 28
>Atlantic City International Airport, NJ 08405
>USA
>Phone:  609-485-4191
>Fax:  609-485-6218
>E-mail: Ben.Willems at faa.gov
>
>Teach CanIt if this mail (ID 2785751) is spam:
>NOTE: This message has been trained as non-spam. If this is wrong, 
>please correct the training as soon as possible.
><http://respite.rpi.edu/b.php?c=s&i=2785751&m=78d39b426ef0>Spam
><http://respite.rpi.edu/b.php?c=n&i=2785751&m=78d39b426ef0>Not spam
><http://respite.rpi.edu/b.php?c=f&i=2785751&m=78d39b426ef0>Forget 
>previous vote
>
>
>_______________________________________________
>ACT-R-users mailing list
>ACT-R-users at act-r.psy.cmu.edu
>http://act-r.psy.cmu.edu/mailman/listinfo/act-r-users


-- 
**Rensselaer**Rensselaer**Rensselaer**Rensselaer**Rensselaer**
Wayne D. Gray; Professor of Cognitive Science
Rensselaer Polytechnic Institute
Carnegie Building (rm 108) ;;for all surface mail & deliveries
110 8th St.; Troy, NY 12180

EMAIL: grayw at rpi.edu, Office: 518-276-3315, Fax: 518-276-3017

for general information see: http://www.rpi.edu/~grayw/

for On-Line publications see: 
http://www.rpi.edu/~grayw/pubs/downloadable_pubs.htm

for the CogWorks Lab see: http://www.cogsci.rpi.edu/cogworks/

If you just have formalisms or a model you are doing "operations 
research" or" AI", if you just have data and a good study you are 
doing "experimental psychology", and if you just have ideas you are 
doing "philosophy" -- it takes all three to do cognitive science.

**Rensselaer**Rensselaer**Rensselaer**Rensselaer**Rensselaer**



More information about the ACT-R-users mailing list