From fjl at rpi.edu Sat Aug 2 20:14:45 2003 From: fjl at rpi.edu (Frank J. Lee) Date: Sat, 02 Aug 2003 20:14:45 -0400 Subject: [ACT-R-users] Hot off the press! Message-ID: !!!STRAIGHT FROM COGNITIVE SCIENCE 2003!!! It was announced at the conference that John is the recipient of this years Rumelhart Prize! All we can say is FINALLY! Let's plan on a big celebration next year at Cognitive Science 2004 in Chicago where he will be honored. -This year's ACT-R Army at Cog Sci 2003- From tkelley at arl.army.mil Mon Aug 4 09:35:04 2003 From: tkelley at arl.army.mil (Kelley, Troy (Civ,ARL/HRED)) Date: Mon, 4 Aug 2003 09:35:04 -0400 Subject: [ACT-R-users] Hot off the press! Message-ID: <01AB5B234641D44EABB3322695F8EAD00A2B53@ARLABML01.DS.ARL.ARMY.MIL> John, Congratulations! The award is very well deserved, and long overdue. Troy Kelley Army Research Laboratory -----Original Message----- From: Frank J. Lee [mailto:fjl at rpi.edu] Sent: Saturday, August 02, 2003 8:15 PM To: act-r-announce at act-r.psy.cmu.edu; act-r-users at act-r.psy.cmu.edu Subject: [ACT-R-users] Hot off the press! !!!STRAIGHT FROM COGNITIVE SCIENCE 2003!!! It was announced at the conference that John is the recipient of this years Rumelhart Prize! All we can say is FINALLY! Let's plan on a big celebration next year at Cognitive Science 2004 in Chicago where he will be honored. -This year's ACT-R Army at Cog Sci 2003- _______________________________________________ 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 From Wolfgang.Schoppek at uni-bayreuth.de Mon Aug 11 08:26:15 2003 From: Wolfgang.Schoppek at uni-bayreuth.de (Wolfgang.Schoppek at uni-bayreuth.de) Date: Mon, 11 Aug 2003 14:26:15 +0200 (MEST) Subject: [ACT-R-users] Production compilation Message-ID: <1084.132.180.58.12.1060604775.squirrel@mail.uni-bayreuth.de> Is there a known bug in the current parameter calculation for new productions (e.g. in the function finish-up)? I get unreasonable low p's for new productions. Hand-calculations according to the tutorial result in p=0.73, ACT-R gives me p=0.09. I get that effect since I've updated my (old ACL) environment to ACT-R 5.0.6. -- Wolfgang Schoppek From taatgen at cmu.edu Mon Aug 11 11:05:11 2003 From: taatgen at cmu.edu (Niels Taatgen) Date: Mon, 11 Aug 2003 11:05:11 -0400 Subject: [ACT-R-users] Production compilation In-Reply-To: <1084.132.180.58.12.1060604775.squirrel@mail.uni-bayreuth.de> Message-ID: <346CA3BE-CC0D-11D7-8B8F-0003930F23B6@cmu.edu> Wolfgang, It all depends on which ACT-R 5.0.6 you have. Probably you have the one that is on the web right now, and it contains an update on how ACT-R handles parameters for new productions (basically it was what I proposed on last year's workshop). You can identify that you have the new code if you have the file "prod-comp-update.lisp" in your ACT-R folder. The change is described in the tutorial unit (unit 9), also available on the website, and in last year's workshop proceedings. What probably explains your low p value, is that in the new system p is initially set to 0, and each time the production is created again, its value is increased so that it converges to the p-value of the parent rule (in your case 0.73). Niels Taatgen On Monday, August 11, 2003, at 08:26 AM, Wolfgang.Schoppek at uni-bayreuth.de wrote: > Is there a known bug in the current parameter calculation for new > productions (e.g. in the function finish-up)? I get unreasonable low > p's > for new productions. Hand-calculations according to the tutorial > result in > p=0.73, ACT-R gives me p=0.09. I get that effect since I've updated my > (old ACL) environment to ACT-R 5.0.6. > > -- Wolfgang Schoppek > =============================================== Niels Taatgen - University of Groningen, Artificial Intelligence Now at: Carnegie Mellon University, Psychology, BH 345E web: http://www.ai.rug.nl/~niels email: taatgen at cmu.edu Telephone: +1 412-268-2815 =============================================== From db30 at andrew.cmu.edu Thu Aug 14 17:09:54 2003 From: db30 at andrew.cmu.edu (db30 at andrew.cmu.edu) Date: Thu, 14 Aug 2003 17:09:54 -0400 Subject: [ACT-R-users] Production compilation equations in unit 9 In-Reply-To: <1084.132.180.58.12.1060604775.squirrel@mail.uni-bayreuth.de> References: <1084.132.180.58.12.1060604775.squirrel@mail.uni-bayreuth.de> Message-ID: <28613534.1060880994@whatever.psy.cmu.edu> To let everyone know, based on the problem Wolfgang reported we discovered that there was an error in the equations for P and C of the learned production in the unit 9 text. That has now been corrected. Dan From abramson at ait.nrl.navy.mil Fri Aug 29 13:41:16 2003 From: abramson at ait.nrl.navy.mil (Myriam Abramson) Date: Fri, 29 Aug 2003 13:41:16 -0400 Subject: [ACT-R-users] act-r production rules development Message-ID: hi! Do you have any tips or pointers to extract act-r production rules from an expert. What should I look for? How should I phrase my questions so that they can articulate the rules themselves? How is it different from knowledge engineering? Thanks for your insights. myriam From rlwest at ccs.carleton.ca Fri Aug 29 20:10:23 2003 From: rlwest at ccs.carleton.ca (Robert West) Date: Fri, 29 Aug 2003 20:10:23 -0400 Subject: [ACT-R-users] act-r production rules development In-Reply-To: Message-ID: <59996F52-DA7E-11D7-812A-000393A479A6@ccs.carleton.ca> I would say that one of the most important issues is whether the task is being driven by procedural or declarative memory. If it is procedural memory then you wouldn't expect the experts to have direct access to the rules they are following (i.e., because we do not have conscious access to procedural memory). Instead their self reports would be based on observing themselves and/or mentally simulating the task and attempting to guess or reason out what the rules must be (although people often mistake this for direct access to the rules). In these cases it is a really good idea to observe them carefully yourself as you will find there are often a lot of things they leave out. If it is driven by declarative memory then they can often report the knowledge they are using while doing the task and you can infer the production rules they used to extract it. Generally speaking, especially in experts, behavior directed just by production rules will be very fast and automatic. So, I would say both observation studies and verbal protocols are important Robert Robert L. West Dept of Psychology Dept of Cognitive Science Carelton University On Friday, August 29, 2003, at 01:41 PM, Myriam Abramson wrote: > > hi! > > Do you have any tips or pointers to extract act-r production rules > from an expert. What should I look for? How should I phrase my > questions so that they can articulate the rules themselves? > How is it different from knowledge engineering? > > Thanks for your insights. > > myriam > > _______________________________________________ > 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 >