From ja+ at cmu.edu Sun Apr 2 12:49:55 2000 From: ja+ at cmu.edu (John Anderson) Date: Sun, 2 Apr 2000 12:49:55 -0400 Subject: post-doc at CMU Message-ID: We have a post-doc available at CMU for a non-CMU PhD with some ACT-R background. Interested candidates can contact me and we can discuss the possibilities. From dario at pathfinder.cbr.com Wed Apr 12 14:54:29 2000 From: dario at pathfinder.cbr.com (Dario Salvucci) Date: Wed, 12 Apr 2000 14:54:29 -0400 Subject: ACT-R multitasking Message-ID: A question about ACT-R and multitasking... We have recently begun an attempt to model driving in ACT-R/PM, and immediately the issue arose of how to handle multiple tasks. There seem to be at least two (related) problems: 1. Interleaving tasks. Since ACT-R is serial, we require some scheduling that allocates pieces of time to each task. One straightforward approach might use several top-level productions that compete through conflict resolution and push task subgoals onto the stack. Each subgoal would do some incremental piece of the task and return control to the higher-level productions. However, this raises a problem of time allocation: the noisy process may happen to allocate large chunks of time to a single process without others getting their share. In addition, there is a question of whether the subtasks should dictate when they complete or whether the scheduler should interrupt subtasks when needed. 2. Prioritizing tasks. In tasks such as driving, certain tasks aren't as important as others (e.g., steering vs. changing radio stations). How can the model allocate time according to task priorities? Again with the straightforward approach above, we could set conflict resolution parameters to values that represent task priorities. However, often these priorities are dynamic and time-related -- for instance, after a steering correction, steering may not be important for another one or two seconds, during which time a lower-priority task could be executed. I'm aware of some work on interleaving for short cognitive tasks (e.g., Byrne's PRP work) and some other work on large-scale ACT-R models of dynamic tasks, but I'm unclear as to how much has been done for multitasking/scheduling of prioritized tasks. Might anyone have any ideas, papers, or experiences related to these issues? Thanks, Dario PS: The EMMA model of eye movements and visual attention introduced at ICCM 2000 is now available at < http://www.cbr.com/~dario/EMMA >. The model is in its alpha version and is probably best suited for serious ACT-R/PM junkies. :) We expect to have a beta version that is fully integrated into ACT-R/PM by the time the ACT-R workshop rolls around. Feel free to email me for more information. --------------------------------------------------------------------------- Dario Salvucci Email: dario at cbr.com Cambridge Basic Research Phone: (617) 374-9669 Four Cambridge Center Fax: (617) 374-9697 Cambridge, MA 02142 USA WWW: http://www.cbr.com/~dario From wschoppe at mason2.gmu.edu Wed Apr 12 15:57:21 2000 From: wschoppe at mason2.gmu.edu (Wolfgang Schoppek) Date: Wed, 12 Apr 2000 15:57:21 -0400 Subject: ACT-R multitasking Message-ID: Hi Dario, it is really funny: at the very moment when your e-mail arrived, I was sitting here, working on a solution for this problem. So let me start with a question. > 1. Interleaving tasks. Since ACT-R is serial, we require some > scheduling that allocates pieces of time to each task. How do you allocate pieces of time to several tasks? Do you work with different G's for different tasks? > One > straightforward approach might use several top-level productions that > compete through conflict resolution and push task subgoals onto the > stack. Each subgoal would do some incremental piece of the task and > return control to the higher-level productions. However, this raises > a problem of time allocation: the noisy process may happen to allocate > large chunks of time to a single process without others getting their > share. In addition, there is a question of whether the subtasks > should dictate when they complete or whether the scheduler should > interrupt subtasks when needed. My solution for that is that I do not use the full depth of the goal stack and use a generic type for most of the goals. When a subgoal comes into play, it is not pushed, but rather, the current goal is rehearsed, tagged with 'stack and replaced by the subgoal. All this is under control of a basic top-level goal. When a goal is popped, the basic goal retrieves the most active goal in memory that has a certain tag, e.g. the 'stack tag. But it could also retrieve a goal with an 'intention tag or with a 'pending tag. Since in my model the sequences under one (sub)goal are quite short (3-8 steps), the model doesn't spend much time between the top-level checks - even if it is actually somewhere deep in a goal hierarchy. Therefore I don't interrupt the ongoing process (although I think that there are events that do interrupt immediately). > 2. Prioritizing tasks. In tasks such as driving, certain tasks aren't > as important as others (e.g., steering vs. changing radio stations). > How can the model allocate time according to task priorities? Again > with the straightforward approach above, we could set conflict > resolution parameters to values that represent task priorities. > However, often these priorities are dynamic and time-related -- for > instance, after a steering correction, steering may not be important > for another one or two seconds, during which time a lower-priority > task could be executed. I also would like to have a solution for that. I think that some of the prioritizing takes place on the symbolic level and can be modeled accordingly. But for strong emergency signals, I think that something happens on lower levels, where e.g. classical conditioning processes take place. The size of deviations from desired values might also be a factor. -- Wolfgang -------------------------------------------------------------------- Dr. Wolfgang Schoppek <<< Tel.: +1 703-993-4663 <<< HUMAN FACTORS & APPLIED COGNITION PROGRAM, George Mason University, Fairfax, VA 22030-4444 http://www.uni-bayreuth.de/departments/psychologie/wolfgang.htm -------------------------------------------------------------------- From mh5r at andrew.cmu.edu Wed Apr 12 18:04:46 2000 From: mh5r at andrew.cmu.edu (Mary I Hart) Date: Wed, 12 Apr 2000 18:04:46 -0400 (EDT) Subject: ACT-R multitasking Message-ID: I think the kind of scheduling, giving time-slices to different processes, and prioritizing that you want to do is discussed extensively in any Operating Systems textbook. I'd start by looking there for an appropriate algorithm which you could then implement in ACT-R. Mary Hart From niels at tcw2.ppsw.rug.nl Thu Apr 13 03:13:18 2000 From: niels at tcw2.ppsw.rug.nl (Niels Taatgen) Date: Thu, 13 Apr 2000 09:13:18 +0200 Subject: Postdoc position in Groningen Message-ID: The department of philosophy has the following position available. It will possibly also involve ACT-R modelling, so I pass it on to the list. Niels Taatgen Postdoc COMPUTATIONAL MODELING OF ABDUCTION Postdoc Position Faculty of Philosophy University of Groningen Department of Philosophy of Science, Logic, and Epistemology Subject of research The first task is to trace, compare, and improve computational models of abductive reasoning that have been developed in logic, philosophy of science, artificial intelligence or cognitive psychology. Subsequent elaboration in general or emphasis in one particular direction are both possible. The research takes place in the context of the Special Program of the Research School in Behavioural and Cognitive Neurosciences (BCN), entitled Computational modeling of structures in performance. Requirements - You are qualified, by PhD or otherwise, in at least one of the following areas: logic, philosophy of science, artificial intelligence, cognitive psychology - You have considerable education in logic and affinity with the three other areas mentioned. - You have finished your PhD - You are able to do research of high quality - You are willing and able to guide an already existing practicum Computational Philosophy of Science - You are able to work harmoniously in a team Offer The University offers a salary depending on education and experience of maximally 7.618 Dutch guilders bruto per month on a fulltime basis. The appointment is for two years, starting as soon as possible. After one year a positive evaluation is required for continuation. The faculty aims at increasing the percentage of female researchers and staff members. Information More information about the position can be obtained from Professor Theo A. F. Kuipers, tel. 0031 (0)50 3636151 (work) or home (0031 (0)50 5263748, e-mail T.A.F.Kuipers at philos.rug.nl Information about the Faculty of Philosophy can be found on www.philos.rug.nl and about BCN on www.bcn.rug.nl Applications Applications (CV, List of Publications, Indication Research Plan) should be send, before May 10, 2000, to the above e-mail address or to Kuipers' post address: Faculty of Philosophy, A-weg 30, 9718 CW Groningen. The Netherlands. From ritter at ist.psu.edu Thu Apr 13 13:29:33 2000 From: ritter at ist.psu.edu (Frank E. Ritter) Date: Thu, 13 Apr 2000 13:29:33 -0400 Subject: ACT-R multitasking Message-ID: There are two interesting Soar models about this problem that you may wish to find. The source code, well, dunno. that's still a problem. Chong's system learned how to sequence tasks and to my eye was quite cool. rchong at soartech.com Aasman's model drove. Not always cited, but worked. Also learned and was compared to data. I think you know him or can recognise him now. j.aasman at research.kpn.com Cheers, =46rank Chong, R. S., & Laird, J. E. (1997). Identifying dual-task executive process knowledge using EPIC-Soar. In Proceedings of the 19th Annual Conference of the Cognitive Science Society. 107-112. Mahwah, NJ: Lawrence Erlbaum. Aasman, J. (1995). Modelling driver behaviour in Soar. Leidschendam, The Netherlands: KPN Research. Aasman, J., & Michon, J. A. (1992). Multitasking in driving. In J. A. Michon & A. Aky=FCrek (Eds.), Soar: A cognitive architecture in perspective. Dordrecht, The Netherlands: Kluwer. =46rank Ritter at ist.psu.edu School of Information Sciences and Technology The Pennsylvania State University University Park, PA 16801-3857 ph. (814) 865-4453 fax (814) 865-5604 http://www.psyc.nott.ac.uk/staff/ritter (archived) http://www.cedcc.psu.edu/ritter/ (temp local one) From bej at sei.cmu.edu Thu Apr 13 14:33:05 2000 From: bej at sei.cmu.edu (Bonnie John) Date: Thu, 13 Apr 2000 14:33:05 -0400 Subject: ACT-R multitasking Message-ID: "Frank E. Ritter" wrote: > > There are two interesting Soar models about this problem that you may wish > to find. The source code, well, dunno. that's still a problem. > > Chong's system learned how to sequence tasks and to my eye was quite cool. > rchong at soartech.com ... > > Chong, R. S., & Laird, J. E. (1997). Identifying dual-task executive > process knowledge using EPIC-Soar. In Proceedings of the 19th Annual > Conference of the Cognitive Science Society. 107-112. Mahwah, NJ: Lawrence > Erlbaum. Dario, Yannick Lallement also wrote a Soar program that learned to do several things at once (through interleaving tasks and learning when to interrupt or give priority to one or the other), on the same task that Ron did, but using a different version of Soar. We compare the two models (and another non-learning one in EPIC) in the 1999 Cognitive Science Conference (I think it was 1999, though it might have been 1998 -- I don't have access to my vita right now). I'm sure Yannick would be happy to give you a pointer to the code but since it involves using Soar and EPIC together, it is probably hard to get to run somewhere other than here. But you might contact him anyway to discuss what we did and how we did it. Yannick Lallement Bonnie From nbrannon at cs.wright.edu Fri Apr 21 09:43:08 2000 From: nbrannon at cs.wright.edu (Nathan Brannon) Date: Fri, 21 Apr 2000 09:43:08 -0400 (EDT) Subject: Question from a novice Message-ID: I've been evaluating a supervisory control task and translating user strategies into production rules (written in plain english). As I was preparing the rules I noticed certain situations where I would normally use the conditional expression IF-THEN-ELSE. In my short relatively short time working with ACT-R, I've only seen a condition and action (IF-THEN). I suspect there is a good reason why this is the case and so I was hoping someone could help me understand alternatives to IF-THEN-ELSE expressions in ACT-R. Although one route is through generating more rules, this seemed to be a more cumbersome approach. Thanks! ________________________________________________________________ Nathan G. Brannon, MSE Ph.D. Candidate Biomedical, Industrial, and Human Factors Engineering Wright State University BIE Dept. / 207 RC Dayton, Ohio 45435 Ph: 937.775.5044 or 937.775.5152 Fax: 937.775.7364 e-mail: nbrannon at cs.wright.edu www: home.earthlink.net/~ngbrannon From Stefani_Nellen at psi-sv2.psi.uni-heidelberg.de Fri Apr 21 10:34:43 2000 From: Stefani_Nellen at psi-sv2.psi.uni-heidelberg.de (Stefani Nellen) Date: Fri, 21 Apr 2000 16:34:43 +0200 Subject: Question from a novice Message-ID: --------------D8B23C2FB67184CDE547B6DB Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Dear Nathan! I'm also a bit of a novice, so I feel sympathetic with you. ad the IF-THEN-Syntax of ACT-R production rules: The reason for this is, I think, the ambition of the ACT community to model the steps of human cognition on the atomic level, i.e. as detailed as possible. A production which contains an IF-THEN-ELSE structure implies that the following cognitive steps have already been made: the criterion for a certain course of action has been identified (IF), and a possible/ plausible/ whatever alternative action as well (ELSE). Basically, this is an entire strategy, compressed in a single production rule. The reason to break down production rules to the simpler IF-THEN is that this constraint enables the modeller to model the gradual formation of strategies, the underlying learning processes and so on. Although you write that to generate more production rules would be "cumbersome" there are, I think some quite nice solutions to the IF-THEN-ELSE problem. You can implement such a stucture in a straightforward way by writing three production rules, the first containing a retieval request for the information: (p search information =goal> isa goal information nil =infomation> content something ==> =goal> information something) and two productions implementing the then/else part, both checking the "something"retrieved in the production before. The first contains in its condition part that the information be your IF, however you specify it. If "something" is this IF, the action occurs. The alternative one is just checking for the information NOT being the special IF you want it to be, via a negation check. In plain english this would be something like: (P altenative 1 IF the goal is (whatever), and there is information of a certain type, THEN action) (p altenative 2 IF the goal is whatever, and there is some information available, which is not of this type, THEN different action) By using the negation check you don't have to specvify altenatives, thus having an equivalent to the if-then-else structure. The formal syntax of this could be: (p alternative1 =goal> isa goal information specified-by-you ==>action!) (p alternative2 =goal> isa goal - information specified by you ==> different action!) There may be more sophisticated ways to do that, especially the larning and modification of such strategies, but unfortunatley I'm under a little bit of time pressure at the moment and can't expand on the topic, maybe some of the experts have some first hand modelling experience for you. I wish you all the best, and have fun with your modelling! Stefani > I've been evaluating a supervisory control task and translating user > strategies into production rules (written in plain english). As I was > preparing the rules I noticed certain situations where I would normally > use the conditional expression IF-THEN-ELSE. In my short relatively short > time working with ACT-R, I've only seen a condition and action > (IF-THEN). I suspect there is a good reason why this is the case and so > I was hoping someone could help me understand alternatives to > IF-THEN-ELSE expressions in ACT-R. Although one route is through > generating more rules, this seemed to be a more cumbersome approach. > > Thanks! > > ________________________________________________________________ > > --------------D8B23C2FB67184CDE547B6DB Content-Type: text/html; charset=us-ascii Content-Transfer-Encoding: 7bit Dear Nathan!
I'm also a bit of a novice, so I feel sympathetic with you.

ad the IF-THEN-Syntax of ACT-R production rules:
The reason for this is, I think, the ambition of the ACT community to model the steps of human cognition on the atomic level, i.e. as detailed as possible. A production which contains an IF-THEN-ELSE structure implies that the following cognitive steps have already been made: the criterion for a certain course of action has been identified (IF), and a possible/ plausible/ whatever alternative action as well (ELSE). Basically, this is an entire strategy, compressed in a single production rule.
The reason to break down production rules to the simpler IF-THEN is that this constraint enables the modeller to model the gradual formation of strategies, the underlying learning processes and so on.

Although you write that to generate more production rules would be "cumbersom e" there are, I think some quite nice solutions to the IF-THEN-ELSE problem.
 You can  implement such a stucture in a straightforward way by writing three production rules, the first containing a retieval request for the information:
(p search information
=goal>
isa goal
information nil
=infomation>
content something
==>
=goal>
information something)

and two productions implementing the then/else part, both checking the "something"retrieved in the production before. The first contains in its condition part that the information be your IF, however you specify it. If "something" is this IF, the action occurs. The alternative one is just checking for the information NOT being the special IF you want it to be, via a negation check.
In plain english this would be something like:
(P altenative 1
IF the goal is (whatever), and there is information of a certain type,
THEN action)
(p altenative 2
IF the goal is whatever, and there is some information available, which is not of this type, THEN different action)

By using the negation check you don't have to specvify altenatives, thus having an equivalent to the if-then-else structure.
The formal syntax of this could be:

(p alternative1
=goal>
isa goal
information specified-by-you
==>action!)

(p alternative2
=goal>
isa goal
- information specified by you
==> different action!)

There may be more sophisticated ways to do that, especially the larning and modification of such strategies, but unfortunatley I'm under a little bit of time pressure at the moment and can't expand on the topic, maybe some of the experts have some first hand modelling experience for you.
I wish you all the best, and have fun with your modelling!
Stefani 

I've been evaluating a supervisory control task and translating user
strategies into production rules (written in plain english).  As I was
preparing the rules I noticed certain situations where I would normally
use the conditional expression IF-THEN-ELSE. In my short relatively short
time working with ACT-R, I've only seen a condition and action
(IF-THEN).  I suspect there is a good reason why this is the case and so
I was hoping someone could help me understand alternatives to
IF-THEN-ELSE expressions in ACT-R.  Although one route is through
generating more rules, this seemed to be a more cumbersome approach.

Thanks! 

________________________________________________________________
--------------D8B23C2FB67184CDE547B6DB-- From rsun at cecs.missouri.edu Fri Apr 21 16:50:45 2000 From: rsun at cecs.missouri.edu (Ron Sun) Date: Fri, 21 Apr 2000 15:50:45 -0500 Subject: recent issues of Cognitive Systems Research Message-ID: Contents of the recent issues of Cognitive Systems Research: ------------------------- Table of Contents for Cognitive Systems Research Volume 1, Issue 1, 1999 Ron Sun, Vasant Honavar and Gregg C. Oden Editorial: Integration of cognitive systems across disciplinary boundaries 1-3 Andy Clark Where brain, body, and world collide 5-17 Arthur M. Glenberg, David A. Robertson, Jennifer L. Jansen and Mina C. Johnson-G lenberg Not Propositions 19-33 Arthur C. Graesser, Katja Wiemer-Hastings, Peter Wiemer-Hastings and Roger Kreuz AutoTutor: A simulation of a human tutor 35-51 Pentti Kanerva Book Review: Artificial Minds, Stan Franklin, MIT Press, Cambridge, MA, 19 95 53-57 Xin Yao Conference Report: Evolutionary computation comes of age 59-64 ------------------------- Table of Contents for Cognitive Systems Research Volume 1, Issue 2, January 2000 Mark H. Bickhard Information and representation in autonomous agents 65-75 Valerie Gray Hardcastle The development of the self 77-86 Umberto Castiello et al. Human inferior parietal cortex `programs' the action class of grasping 89-97 Marsha C. Lovett, Larry Z. Daily and Lynne M. Reder A source activation theory of working memory: cross-task prediction of perf ormance in ACT-R 99-118 A. El Imrani, A. Bouroumi, H. Zine El Abidine, M. Limouri and A. Essaod A fuzzy clustering-based niching approach to multimodal function optimizati on 119-133 ------------------------- Table of Contents for Cognitive Systems Research Volume 1, Issue 3, April 2000 Brijesh Verma and Chris Lane Vertical jump height prediction using EMG characteristics and neural networ ks 135-141 Scott A. Huettel and Gregory Lockhead Psychologically rational choice: selection between alternatives in a multip le-equilibrium game 143-160 Robert C. Mathews, Lewis G. Roussel, Barbara P. Cochran, Ann E. Cook and Deborah L. Dunaway The role of implicit learning in the acquisition of generative knowledge 161-174 ------------------------------------------------------------------- Publish your work with Cognitive Systems Research --- the new journal devoted to the interdisciplinary study of cognitive science http://www.elsevier.nl/locate/cogsys Elsevier Science Co-Editors-in-Chief Ron Sun. E-mail: rsun at cecs.missouri.edu Vasant Honavar. E-mail: honavar at cs.iastate.edu Gregg Oden. E-mail: gregg-oden at uiowa.edu Cognitive Systems Research covers all topics of cognition, including ' Perception ' Memory ' Learning ' Action and Behavior ' Problem-Solving and Cognitive Skills ' Knowledge Representation and Reasoning ' Language and Communication ' Agents ' Integrative and Interdisciplinary Studies For a full description of subjects and submission information, access the Website: http://www.elsevier.nl/locate/cogsys or http://www.cecs.missouri.edu/~rsun/journal.html ------------------------------------------------------------------- =========================================================================== Prof. Ron Sun http://www.cecs.missouri.edu/~rsun CECS Department phone: (573) 884-7662 University of Missouri-Columbia fax: (573) 882 8318 201 Engineering Building West Columbia, MO 65211-2060 email: rsun at cecs.missouri.edu http://www.cecs.missouri.edu/~rsun http://www.cecs.missouri.edu/~rsun/journal.html http://www.cecs.missouri.edu/~rsun/clarion.html http://www.elsevier.com/locate/cogsys =========================================================================== From trafton at itd.nrl.navy.mil Mon Apr 24 11:38:20 2000 From: trafton at itd.nrl.navy.mil (Greg Trafton) Date: Mon, 24 Apr 2000 11:38:20 -0400 (EDT) Subject: NRL postdoc Message-ID: (We're looking for a postdoc to help with a new project. NRL is a great place to be...) Postdoc Position Naval Research Laboratory Washington, DC NRL is currently working on a project to combine computational cognitive architectures (e.g., ACT-R) with machine learning/genetic algorithms in novel ways. We are investigating how combining these two different computational systems will lead to more robust cognitive models with better performance. The project has two main goals. The first goal is to add cognitive representation to a machine learning system. The second goal is to integrate the workings of both the cognitive architecture and the machine learning system. This project will use a number of different methodologies including protocol analysis, psychological experiments, cognitive modeling, and machine learning experiments. Initial responsibilities will be to create a computational cognitive model of a simple task (using ACT-R), create a similar model using an in-house machine learning/GA system, and combine them to use the best of both systems. Additional responsibilites will be to empirically evaluate different versions of the system. Background or experience with either a computational cognitive architecture (ACT-R, Soar, EPIC, etc.) or experience using machine learning techniques is necessary. We do not expect the candidate to know both a computational cognitive architecture and a machine learning technique, though we do expect the candidate to be willing to learn the "other" system. We expect that the successful candidate will have a Ph.D. in Computer Science or Psychology but Ph.D.'s in other areas can be considered. Programming experience is a must. The appointment would be for one year; an additional two years would be contingent on availability of funds and performance. We currently have funding for this project for three years. This appointment would be either an ASEE or NRC postdoc position; only American citizens are eligible. Post Doc Stipend: $47,000/year for first year Please send a vita, a statement of research interest, and a cover letter explaining why you wish to be considered for this position to either Greg Trafton at trafton at itd.nrl.navy.mil or phone 202-767-3479 or fax number 202-404-4080 or snail mail: Naval Research Lab, Code 5513 4555 Overlook Av. S.W. Washington, DC 20375-5337 or to Alan Schultz at schultz at aic.nrl.navy.mil or phone 202-767-2684 or fax number 202-767-3172 or snail mail: Navy Center for Applied Research in Artificial Intelligence Naval Research Laboratory, Code 5515 4555 Overlook Ave. S.W. Washington DC 20375-5337 From tkelley at arl.mil Fri Apr 28 14:32:04 2000 From: tkelley at arl.mil (Troy Kelley) Date: Fri, 28 Apr 2000 14:32:04 -0400 Subject: ACT-R and Neural Nets Message-ID: Hello Everyone, I am attempting to tie neural nets to ACT-R (and perhaps SOAR) for a project I am working on called MINDSS (Modeling and Integration of Neurological Dynamics with Symbolic Structures). The project is becoming very high profile and I have access to our supercomputers here at the Army Research Laboratory to do the MINDSS work. Anyway, my questions are: 1) As anyone out there had any experience with NeuralWare Software? 2) Can anyone recommend general neural net software that generates portable code (C or LISP)? 3) Are there "hooks" in ACT-R that would let me link a neural network to it and is there documentation for this? 4) Am I crazy? Thanks, Troy Kelley Army Research Laboratory From CHIPMAS at ONR.NAVY.MIL Fri Apr 28 15:28:45 2000 From: CHIPMAS at ONR.NAVY.MIL (Chipman, Susan) Date: Fri, 28 Apr 2000 15:28:45 -0400 Subject: ACT-R and Neural Nets Message-ID: Sandra Marshall and her associates at SDSU have been doing this under support from ONR's special program (now ended), Hybrid Architectures for Complex Learning. smarshall at sciences.sdsu.edu Susan F. Chipman, Ph.D. Office of Naval Research, Code 342 800 N. Quincy Street Arlington, VA 22217-5660 phone: 703-696-4318 Fax: 703-696-1212 -----Original Message----- From: Troy Kelley [mailto:tkelley at arl.mil] Sent: Friday, April 28, 2000 2:32 PM To: act-r-users at andrew.cmu.edu Subject: ACT-R and Neural Nets Hello Everyone, I am attempting to tie neural nets to ACT-R (and perhaps SOAR) for a project I am working on called MINDSS (Modeling and Integration of Neurological Dynamics with Symbolic Structures). The project is becoming very high profile and I have access to our supercomputers here at the Army Research Laboratory to do the MINDSS work. Anyway, my questions are: 1) As anyone out there had any experience with NeuralWare Software? 2) Can anyone recommend general neural net software that generates portable code (C or LISP)? 3) Are there "hooks" in ACT-R that would let me link a neural network to it and is there documentation for this? 4) Am I crazy? Thanks, Troy Kelley Army Research Laboratory