Learning facts in ACT

[Wayne Gray]

Interesting question. When I was a graduate student, my mentor Leo 
Postman was a verbal learner. Such problems as Richard raises were 
the meat and potatoes of verbal learning. But the verbal learners 
quit the field during the 80's and no one took up their puzzles -- 
until now maybe. Clearly, what was lacking in those old theories was 
a mechanistic account that could be rigorously specified.

So, first of all, most paired-associates (PAs) used in experiments 
have low pre-experimental association. Hence, the first mystery is 
how do the new and usually arbitrary associations get learned and 
used over the older and more established associations? Especially 
given the fact that learning pairs such as "locomotive-nurse" 
interferes minimally with pre-experimental associations such as 
"locomotive-black."

I believe that Richard's essential insight is correct, the simple 
actr theories need to be regarded as teaching tools, and not as 
theories of PA learning.

An answer to Richard's question is contained in models of switch cost 
for serial attention that Erik Altmann has been pursuing. Follow this 
link:

http://www.hfac.gmu.edu/People/altmann/pubs.html

Your best bet would be to download the three "manuscript submitted" 
papers. You might also be interested in glancing at the two Altmann 
and Gray (1999; 1998) Cognitive Science Conference papers as well.

If the PA is A-B, when a Ss see's "A" what gets retrieved is not "B" 
but an episodic instruction of what to do with "A" NOW. This 
instruction might contain the pair AB, but whatever it contains, 
essentially it tells you to respond "B" NOW. It is this instruction 
(let's call it AB) that gets retrieved and strengthened. If you 
accidently retrieve an AC instruction, this will cause you to respond 
"C." However, you can overcome this response by allowing AC  to decay 
while you massively rehearse AB.

In the Altmann and Gray models we estimated that one rehearsal can 
occur every 100 msec. Hence, when the A-B pair is on the screen and 
the Ss has just say "C", the Ss can rehearse A-B about 10 times a 
second while ignoring C. The strength of C can decay while that of B 
increases. As memory returns the most active trace, for the most part 
it should return the most recently rehearsed trace, in this case the 
instruction AB that results in the Ss responding "B."

Note that this approach leaves a lot of room for Proactive 
Interference. After you stop actively rehearsing B its strength 
decays rapidly with the consequence that the difference between its 
strength and C's strength lessens. With a few assumptions (built into 
actr) regarding noise, overtime you are likely to falsely retrieve C. 
Indeed, if C were on a first list of PAs and had been well learned, 
whereas B was on a second list of PAs and had been rehearsed only a 
few times, actr might well predict that the strength of the AB 
instruction would decay to be less than the strength of the AC 
instruction, with the result that (after an appropriate retention 
interval) AC would be returned rather than AB.

I believe this approach can be just as easily applied to model 
retroactive interference.

Anyone interested in pursuing this should read Postman's masterful 
summary of the verbal learning literature:

Postman, L. (1971). Transfer, interference, and forgetting. In J. W. 
Kling & L. A. Riggs (Eds.), Woodworth & Scholsberg's Experimental 
Psychology (3rd ed., pp. 1019-1132). New York: Holt, Rinehart, and 
Winston, Inc.

I will also throw in the following reference, as I used to be quite fond of it:

Postman, L., & Gray, W. D. (1977). Maintenance of prior associations 
and proactive inhibition. Journal of Experimental Psychology:  Human 
Learning and Memory, 3, 255-263.

Cheers,

Wayne

>
>At 6:23 PM +0000 1/25/00, Richard M Young wrote:
>>ACTors:
>>
>>I have an embarrassingly simple question, or set of related questions,
>>about fact-learning in ACT.  For the purpose of clarity, I'll pose the
>>question in the context of learning a set of paired-associates, although I
>>think the point is more general.  I suspect the answer already exists in a
>>model somewhere, and I just need to be pointed to it.
>>
>>Let's take as a starting point the (obviously over-simplified) model of
>>paired-associate learning and retrieval in the file "paired associate" in
>>Unit 6 of the ACT tutorial.  The crucial part is two rules (there are only
>>three anyway), one of which retrieves the "pair" if it can, and if it can't
>>the other comes into play and "studies" the pair as it is presented.  As is
>>pointed out in the tutorial, the retrieval rule serves to reinforce the
>>activation of the pair twice, once because it is retrieved on the LHS of
>>the rule, and once more when the pair is re-formed from being popped from
>>the goalstack on the RHS.  Notice that "studying" only boosts the
>>activation of the pair once, when it is formed (or re-formed) on the RHS.
>>
>>I got to wondering what would happen if the modelled S ever got into
>>its/his/her head an INCORRECT pair, i.e. with a valid stimulus paired with
>>an incorrect response.  As the model stands, the error would never be
>>corrected, because the erroneous chunk would repeatedly be retrieved, and
>>would be reinforced (twice) each time.  However, it is probably unrealistic
>>to suppose that S doesn't read the feedback just because a response has
>>been retrieved, so there is the opportunity to notice that the retrieved
>>response is wrong and to "study" the correct response.  However, each time
>>that happens, the erroneous chunk gets reinforced twice but the correct
>>chunk only once, as we have seen.  So, given that the erroneous chunks
>>starts off more active than the correct one, except for a vanishingly low
>>probability sequence of events, the correct chunk would never get learned
>>to the point of being retrieved.
>>
>>OK, so it's a crazily over-simplified model, but it does raise the question
>>of how *would* ACT learn paired associates given that it starts off with,
>>or at any stage acquires, erroneous pairs?  I've thought of a couple of
>>ways, but I'm not even sure they'd really work, and they certainly don't
>>seem like convincing stories:
>>
>>(1) Because a retrieval is not guaranteed to be correct, it should not
>>automatically be popped on the RHS of a retrieval rule.  If the model waits
>>for feedback and makes sure it pops only a correct pair, then a correct
>>chunk will be reinforced (once) on each trial.  Unfortunately, the
>>erroneous chunk also gets reinforced once, by being retrieved on the LHS.
>>Because the correct chunk is reinforced AFTER the erroneous one, it profits
>>from recency, and I suppose it's possible that with patience and some luck
>>with the noise, on some occasion the two chunks will be close enough in
>>activation that the correct pair gets retrieved and therefore twice
>>reinforced, and thereafter is likely to win.  But the story doesn't sound
>>convincing.  (And solutions which involve the repeated, deliberate,
>>multiple rehearsal of the correct chunk sound too contrived.)
>>
>>(2) When an erroneous retrieval occurs, and the model discovers that it's
>>wrong from the feedback, as well as learning a correct pair it could also
>>learn the incorrect pair with an additional annotation (attribute) of
>>"wrong".  The retrieval would need to become more elaborate: after
>>retrieving a pair in reply to a probe with the stimulus, the model would
>>check whether it could also retrieve an extended pair marked wrong using
>>the stimulus and the retrieved response.  If it couldn't, OK.  If it could,
>>then it would need to retrieve another pair, with the previous response
>>explicitly negated.  (I think that's possible).  Well, maybe, but again it
>>seems rather contrived.
>>
>>Can anyone tell me how this is done better?
>>
>>-- Richard
>
>======================================================
>   Christian Schunn           Applied Cognitive Program
>   Psychology 3F5             cschunn at gmu.edu
>   George Mason University    (703)-993-1744  Voice
>   Fairfax, VA 22030-4444     (703)-993-1330  Fax
>   http://www.hfac.gmu.edu/~schunn
>======================================================

_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/
Wayne D. Gray  HUMAN FACTORS & APPLIED COGNITIVE PROGRAM

SNAIL-MAIL ADDRESS (FedX et al)     VOICE: +1 (703) 993-1357
George Mason University               FAX: +1 (703) 993-1330
ARCH Lab/HFAC Program                           *********************
MSN 3f5                                         * Work is infinite, *
Fairfax, VA  22030-4444                         * time is finite,   *
http://hfac.gmu.edu                             * plan accordingly. *
_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/