From D.Brumby at cs.ucl.ac.uk Tue Sep 1 11:41:15 2009 From: D.Brumby at cs.ucl.ac.uk (Duncan Brumby) Date: Tue, 1 Sep 2009 16:41:15 +0100 Subject: [ACT-R-users] Research Position Message-ID: We are looking to hire a Research Assistant to join the Interactions on the Move project at the UCL Interaction Centre. The RA will be responsible for the running of experiments aimed at investigating how people perform multiple ongoing tasks while driving. Candidates should have a degree in Psychology (or equivalent subject) and have experience of running controlled experiments with human subjects. Effective working knowledge of statistical data analysis tools is essential (preferably R). Working knowledge of JAVA is desirable. This post is flexible in terms of working hours. It can be either a full- time position for one year starting November 2009, or a half-time position for two years. Closing date for applications is 1st October 2009. This is an exciting opportunity to join a vibrant HCI group at London's leading research university. For more information please see: http://bit.ly/1gYFO9 ________________________________ Dr. Duncan Brumby UCL Interaction Centre University College London Gower Street London WC1E 6BT UK +44 (0)20 7679 0689 brumby at cs.ucl.ac.uk www.uclic.ucl.ac.uk/people/d.brumby/ -------------- next part -------------- An HTML attachment was scrubbed... URL: From bej at cs.cmu.edu Tue Sep 8 08:56:19 2009 From: bej at cs.cmu.edu (Bonnie John) Date: Tue, 08 Sep 2009 08:56:19 -0400 Subject: [ACT-R-users] [Fwd: TODAY: Dissertation Proposal: Leonghwee Teo, Tuesday, September 8th, 3:30pm (NSH 1305)] Message-ID: <4AA65473.3050400@cs.cmu.edu> Sorry I forgot to ask Leonghwee to send this out to ACT-R. My bad. Hope to see some of you there - ACT-R underlies all his work, specifically, it is based on, and expands, SNIF-ACT. Bonnie -------- Original Message -------- Subject: TODAY: Dissertation Proposal: Leonghwee Teo, Tuesday, September 8th, 3:30pm (NSH 1305) Date: Tue, 8 Sep 2009 00:35:32 -0400 From: Leonghwee Teo To: hcii-members at cs.cmu.edu DISSERTATION PROPOSAL Modeling Goal-Directed User Exploration in Human-Computer Interaction Leonghwee Teo Human Computer Interaction Institute Tuesday, September 8th, 3:30pm Newell Simon Hall 1305 COMMITTEE Bonnie E. John (Chair) Aniket Kittur Brad A. Myers Peter L. Pirolli (PARC) ABSTRACT Designing user-interfaces so that first-time or infrequent users can accomplish their goals by exploration has been an enduring challenge in Human-Computer Interaction (HCI). Iterative user-testing is an effective but costly method to develop user-interfaces that support use through exploration. A complementary method is to use modeling tools that can generate predictions of user exploration given a user-interface and a goal description. Recent computational models of goal-directed user exploration have focused on predicting user exploration of websites and demonstrated how predictions can inform user-interface design. These models employ the common concepts of label following and information scent, that the user?s choice is partly determined by the semantic relevance between the user?s goal and the options presented in the user-interface. However, in addition to information scent, other factors including the layout position and grouping of options in the user-interface also affect user exploration and the likelihood of success. In this dissertation, I propose research work for a modeling tool called CogTool-Explorer. CogTool-Explorer will consider the information scent, the visual search process, and the layout and grouping of options in the user-interface in concert, to make more accurate predictions of goal-directed user exploration compared to prior modeling tools. The proposed research will focus on the modeling and predictions by CogTool-Explorer, with a secondary focus on making CogTool-Explorer into a tool for practitioner use. In the presentation, I will first review and synthesize the related work on goal-directed user exploration in HCI and highlight the research gaps that are the focus of this dissertation. I will then present the work I have done to address some of these gaps, and propose the remaining work to complete this dissertation. DOWNLOADS Proposal document: http://www.andrew.cmu.edu/user/teo/Teo_Dissertation_Proposal.pdf Proposal flyer: http://www.andrew.cmu.edu/user/teo/Teo_Proposal_Flyer.pdf Hope to see you there! Leonghwee From rickl at umich.edu Wed Sep 9 08:16:14 2009 From: rickl at umich.edu (Richard L. Lewis) Date: Wed, 9 Sep 2009 08:16:14 -0400 Subject: [ACT-R-users] Postdoc position: reinforcement learning and long-lived agents Message-ID: <40417025-7426-4E84-B5FD-879E06A7FD9C@umich.edu> Postdoctoral fellow position (reinforcement learning) As part of a just funded NSF project, we are looking for a postdoctoral fellow to do research on extending reinforcement learning ideas into an architecture for building long-lived agents. A number of research issues including learning good representations and models, using internal rewards, control of internal as well as external actions, abstraction of actions into skills, and the use of episodic and semantic memories within reinforcement learning are in the scope of the project. A mix of empirical and theoretical work is called for in the project which funds Satinder Singh, John Laird, Richard Lewis and Thad Polk, all at the University of Michigan. The ideal candidate would have done a thesis on a reinforcement learning topic. If you are interested in applying for such a position, please send your details to baveja at umich.edu (please don't spam the whole mailing list). The start date can be as early as Fall 2009. -------------------------- Richard L. Lewis rickl at umich.edu Professor http://www-personal.umich.edu/~rickl/ Department of Psychology Voice: (734) 763-1466 University of Michigan Fax: (734) 763-7480 530 Church Street Office: East Hall 4428F Ann Arbor, MI 48109-1043 -------------- next part -------------- An HTML attachment was scrubbed... URL: From niels at ai.rug.nl Wed Sep 9 08:59:22 2009 From: niels at ai.rug.nl (Niels Taatgen) Date: Wed, 9 Sep 2009 14:59:22 +0200 Subject: [ACT-R-users] PhD position in Groningen Message-ID: <03EE7173-F24E-472F-A922-80F02797DF95@ai.rug.nl> 4 year PhD position in Cognitive Modeling - University of Groningen Cognitive models as believable game opponents The goal of the project is to create cognitive models of how people play games, and use such models to create computer opponents that behave like human opponents. The main challenge of computer (AI) opponents in games has most often been to create an opponent that is smart enough to be able to compete with human players. However, as the classical case of chess illustrates, this does not necessarily mean that the computer plays like a human. Creating an opponent that plays like a human has the advantage that it creates a much richer game experience, especially if the computer player is capable of learning, and can adapt its expertise to the expertise of the player. A pilot project with the game of Set (http://www.ai.rug.nl/~niels/set-app/index.html ) has already demonstrated these virtues. From a scientific point of view, studying human cognition in the context of games offers broad opportunities to gain insights in how players reason about games and their opponents. The project will consist of experimental work with human subjects to gain insight in human performance and learning in games, using eye- tracking to collect fine-grained data, cognitive modeling using the ACT-R cognitive architecture, and construction of computer game opponents. Earlier work on the Set game will serve as a starting point of the project, and will be extended to other games as the project progresses. Environment The project will be supervised by Prof. dr. Niels Taatgen (http://www.ai.rug.nl/~niels/ ), and will be carried out in the Cognitive Modeling group (http://research.ai.rug.nl/index.php/cogmod ) of the Artificial Intelligence department of the University of Groningen. A monthly stipend will be provided of approximately 1400 Euro. The four-year Ph.D project will be carried out within the school of behavioral and cognitive neuroscience. Assistance with stay permits in the Netherlands for non- EU citizens will be provided by the university. Requirements We are looking for a candidate with a degree in Cognitive Science, Artificial Intelligence, Cognitive Psychology or Computer Science with experience in programming and a keen interest in psychological processes. In addition, it is advantageous to have experience with cognitive modeling. Applications Write your application, together with a curriculum vitae and the names of two references as (preferably PDF) attachment by e-mail to n.a.taatgen at rug.nl On November 2, the received applications will be examined, but applications arriving after that date may still be considered, until a suitable candidate has been found. Information about the scholarship, work content etc may be provided by Prof.dr. N.A. Taatgen, +31 50 3636435 or email n.a.taatgen at rug.nl). =============================================== Niels Taatgen - Professor University of Groningen, Artificial Intelligence web: http://www.ai.rug.nl/~niels email: niels at ai.rug.nl Telephone: +31 50 3636435 =============================================== From Jerry.Ball at mesa.afmc.af.mil Mon Sep 14 13:45:57 2009 From: Jerry.Ball at mesa.afmc.af.mil (Ball, Jerry T Civ USAF AFMC 711 HPW/RHAC) Date: Mon, 14 Sep 2009 13:45:57 -0400 Subject: [ACT-R-users] Need for Multi-Level Activation Spread in ACT-R Message-ID: We are in the process of mapping the linguistics representations that are generated by our language comprehension model into a situation model based semantic representation. We are trying to do this in a representationally reasonable way within the ACT-R architecture. The problem we face is the many-to-many mapping between words and concepts. Individual words may map to multiple concepts, and individual concepts may may to multiple words. Given this many-to-many mapping, we would like to use mapping chunks to map from words to concepts. The mapping chunks would encode a single mapping relationship (e.g. a separate mapping chunk to map from the word "bank" to the financial institution concept; from the word "bank" to the river bank concept; from the concept dog to the word "dog"; from the concept dog to the word "canine"). When processing a word, the goal is to retrieve the contextually relevant concept. We would like to accomplish this in a single retrieval, however, we do not know how to do this given the single-level activation spreading mechanism in ACT-R. Since there is no direct link between a word and a concept if mapping chunks are used (i.e. there is no slot in the concept that contains the word), the word will not spread activation to the concept. Instead, given the use of mapping chunks, it appears that two retrievals are needed: 1) given the word, retrieve a mapping chunk, and 2) given a mapping chunk, retrieve a concept. Since our model of language comprehension is already slower than humans at processing language, any extra retrievals are problematic. In fact, we have already eliminated an extra retrieval in determining the part-of-speech of a word. Previously, two retrievals were needed: 1) retrieve the word corresponding to the perceptual input, and 2) given the word (and context) retrieve the part-of-speech of the word. While we were successful in eliminating a retrieval, the resulting word-pos chunks contain a mixture of word form information (e.g. the letters and trigrams in the word) and pos information. Even so, they do not yet contain any representation of phonetic, phonemic, syllabic or morphemic information. With just letter and trigram information, long words contain many slots. Ideally, we would like to represent letter and trigram information independently of each other and POS information (allowing them to interact in retrieving a word), but given the single-level activation spreading mechanism in ACT-R doing so would necessitate multiple independent retrievals, which would fail to capture the interaction of letter and trigram information that leads to successful retrievals of words in the face of variability in the perceptual form (e.g. "arispeed" should retrieve "airspeed"). The fall back for mapping words to concepts is to embed all the possible concepts as slot values in a word and vice versa. While we consider this a representationally problematic solution -- word and concept chunks will wind up needing many extra slots, we do not know how else to work around the single-level activation spread in ACT-R. The primary empirical argument against the need for multi-level activation spread in ACT-R is based on studies which show no activation from words like "bull" to words like "milk", even though "bull" activates "cow" and "cow" activates "milk". Even if it is true that there are no instances of "indirect" activation from "bull" to "milk", this does not rule out the need for multi-level activation spread. There is a hidden assumption that "cow" and "bull" are directly associated, and that "cow" and "milk" are also directly associated. Such direct associations may seem reasonable in small-scale models addressing specific spreading activation phenomena, but they are questionable in a larger-scale model. Do we really want to include all the direct associates of "cow" as slot values in the "cow" chunk, and do the same for all other chunks? We understand that the inclusion of a multi-level activation spreading mechanism in ACT-R would be computationally explosive. However, we would like to have the capability to explore use of such a mechanism and to look for ways to keep it computationally tractable. We have already dealt with the problem of computational explosion in our word retrieval mechanism. Originally, we attempted to use a "soft constraint" retrieval mechanism for words. All words in DM were candidates for retrieval--the most highly activated word being retrieved. With just 2500 words in DM, the activation calculations slowed the model down considerably. To manage retrievals in a tractable manner we implemented a disjunctive retrieval capability combined with a new perceptual span mechanism -- the model first tries a hard-constraint retrieval on the entire perceptual span (which is larger than a word) using the "get-chunk" function (and chop-string under the covers). If get-chunk succeeds (indicating that there is a chunk in DM corresponding to the entire perceptual span) a retrieval is constructed using the entire perceptual span as a hard constraint to retrieve the corresponding multi-word unit in DM, if this fails, the model backs-off and uses the first space delimited word (using chop-string) in the perceptual span to check for a corresponding word in DM -- if a match is found with get-chunk, a retrieval is constructed to retrieve the word. If all else fails, we construct a retrieval that imposes a hard constraint on the first letter (this is less than ideal, but a reasonable compromise). The overall effect is a (nearly) soft-constraint retrieval implemented in a computationally tractable way. A similar capability to effect multi-level activation spread in a computationally tractable manner would be highly desirable. Jerry -------------- next part -------------- An HTML attachment was scrubbed... URL: From db30 at andrew.cmu.edu Mon Sep 14 15:13:12 2009 From: db30 at andrew.cmu.edu (db30 at andrew.cmu.edu) Date: Mon, 14 Sep 2009 15:13:12 -0400 Subject: [ACT-R-users] Need for Multi-Level Activation Spread in ACT-R In-Reply-To: <8987_1252950393_n8EHkWM5019203_E2C53FE31589534B9B15CD4B996DFA9C0434D94C@VFOHMLAO01.Enterprise.afmc.ds.af.mil> References: <8987_1252950393_n8EHkWM5019203_E2C53FE31589534B9B15CD4B996DFA9C 0434D94C@VFOHMLAO01.Enterprise.afmc.ds.af.mil> Message-ID: <187EA6A71B4EB1D629AFFAA1@act-r6.cmu.edu> I'm not sure exactly what you're looking for, but are the existing hooks for the spreading activation mechanism insufficient for trying out something else? There is no constraint that chunks must have a direct association for there to be an Sji between them. They only get one by default if there is, but one can be set explicitly between arbitrary chunks using the set-sji command or by using the :sji-hook if you want to compute the values on the fly instead of in advance. If that doesn't work, then you can also just replace the entire spreading activation mechanism by using the :spreading-hook parameter if you want something completely different. That hook will short circuit the whole spreading computation and let you return any value to add into the activation of the chunk without affecting the other components of the activation equation. To go along with the declarative hooks, it's also possible to add extra parameters to chunks to hold any additional information you might need with the extend-chunks command. So, you could keep tables of connections or other data you need without having to store it in slots of chunks. I think those hooks and commands provide a fair amount of flexibility for experimenting with new or modified mechanisms, but if there's something else you feel is necessary let me know. Dan --On Monday, September 14, 2009 1:45 PM -0400 "Ball, Jerry T Civ USAF AFMC 711 HPW/RHAC" wrote: > > > We are in the process of mapping the linguistics representations that are > generated by our language comprehension model into a situation model > based semantic representation. We are trying to do this in a > representationally reasonable way within the ACT-R architecture. The > problem we face is the many-to-many mapping between words and concepts. > Individual words may map to multiple concepts, and individual concepts > may may to multiple words. Given this many-to-many mapping, we would like > to use mapping chunks to map from words to concepts. The mapping chunks > would encode a single mapping relationship (e.g. a separate mapping chunk > to map from the word "bank" to the financial institution concept; from > the word "bank" to the river bank concept; from the concept dog to the > word "dog"; from the concept dog to the word "canine"). When processing a > word, the goal is to retrieve the contextually relevant concept. We would > like to accomplish this in a single retrieval, however, we do not know > how to do this given the single-level activation spreading mechanism in > ACT-R. Since there is no direct link between a word and a concept if > mapping chunks are used (i.e. there is no slot in the concept that > contains the word), the word will not spread activation to the concept. > Instead, given the use of mapping chunks, it appears that two retrievals > are needed: 1) given the word, retrieve a mapping chunk, and 2) given a > mapping chunk, retrieve a concept. Since our model of language > comprehension is already slower than humans at processing language, any > extra retrievals are problematic. In fact, we have already eliminated an > extra retrieval in determining the part-of-speech of a word. Previously, > two retrievals were needed: 1) retrieve the word corresponding to the > perceptual input, and 2) given the word (and context) retrieve the > part-of-speech of the word. While we were successful in eliminating a > retrieval, the resulting word-pos chunks contain a mixture of word form > information (e.g. the letters and trigrams in the word) and pos > information. Even so, they do not yet contain any representation of > phonetic, phonemic, syllabic or morphemic information. With just letter > and trigram information, long words contain many slots. Ideally, we would > like to represent letter and trigram information independently of each > other and POS information (allowing them to interact in retrieving a > word), but given the single-level activation spreading mechanism in ACT-R > doing so would necessitate multiple independent retrievals, which would > fail to capture the interaction of letter and trigram information that > leads to successful retrievals of words in the face of variability in the > perceptual form (e.g. "arispeed" should retrieve "airspeed"). > > > > The fall back for mapping words to concepts is to embed all the > possible concepts as slot values in a word and vice versa. While we > consider this a representationally problematic solution -- word and > concept chunks will wind up needing many extra slots, we do not know how > else to work around the single-level activation spread in ACT-R. > > > > The primary empirical argument against the need for multi-level > activation spread in ACT-R is based on studies which show no activation > from words like "bull" to words like "milk", even though "bull" activates > "cow" and "cow" activates "milk". Even if it is true that there are no > instances of "indirect" activation from "bull" to "milk", this does not > rule out the need for multi-level activation spread. There is a hidden > assumption that "cow" and "bull" are directly associated, and that "cow" > and "milk" are also directly associated. Such direct associations may > seem reasonable in small-scale models addressing specific spreading > activation phenomena, but they are questionable in a larger-scale model. > Do we really want to include all the direct associates of "cow" as slot > values in the "cow" chunk, and do the same for all other chunks? > > > > We understand that the inclusion of a multi-level activation > spreading mechanism in ACT-R would be computationally explosive. However, > we would like to have the capability to explore use of such a mechanism > and to look for ways to keep it computationally tractable. We have > already dealt with the problem of computational explosion in our word > retrieval mechanism. Originally, we attempted to use a "soft constraint" > retrieval mechanism for words. All words in DM were candidates for > retrieval--the most highly activated word being retrieved. With just 2500 > words in DM, the activation calculations slowed the model down > considerably. To manage retrievals in a tractable manner we implemented a > disjunctive retrieval capability combined with a new perceptual span > mechanism -- the model first tries a hard-constraint retrieval on the > entire perceptual span (which is larger than a word) using the > "get-chunk" function (and chop-string under the covers). If get-chunk > succeeds (indicating that there is a chunk in DM corresponding to the > entire perceptual span) a retrieval is constructed using the entire > perceptual span as a hard constraint to retrieve the corresponding > multi-word unit in DM, if this fails, the model backs-off and uses the > first space delimited word (using chop-string) in the perceptual span to > check for a corresponding word in DM -- if a match is found with > get-chunk, a retrieval is constructed to retrieve the word. If all else > fails, we construct a retrieval that imposes a hard constraint on the > first letter (this is less than ideal, but a reasonable compromise). The > overall effect is a (nearly) soft-constraint retrieval implemented in a > computationally tractable way. > > > > A similar capability to effect multi-level activation spread in a > computationally tractable manner would be highly desirable. > > > > Jerry > > From cl at cmu.edu Tue Sep 15 12:52:24 2009 From: cl at cmu.edu (Christian Lebiere) Date: Tue, 15 Sep 2009 12:52:24 -0400 Subject: [ACT-R-users] Special Issue CFP Message-ID: ? Call For Papers Journal of Artificial General Intelligence Special Issue on Model Comparison for Cognitive Architectures and AGI The purpose of this special issue is to explore the merits of a comparative approach for understanding Artificial General Intelligent (AGI) systems. ?This approach is common in the field of cognitive modeling, where different theories of cognition are instantiated as computational architectures and applied to common tasks to establish to their respective scope and limits. Within the field of cognitive modeling, comparison efforts have been recognized as crucial for making scientific progress, and the method is now finding its way into a number of related fields in cognitive science. ?But model comparison need not be viewed only as a theoretical exercise; in fact, the drive to implement unified theories of cognition as computational architectures and test them against a range of human performance data in dynamic, complex and potentially ill-structured task environments is, at root, no different than the call to develop AI systems that can generalize beyond narrow, task-specific applications. ?In this light, we view model comparison as a means to advance both cognitive science and the study of AGI systems and to reconcile traditions that historically emerged as complementary but have since evolved, for all practical purposes, as independent disciplines. ? The structure and content of this special issue are influenced by a particular model comparison challenge recently organized to explore a generic dynamic decision making task, the Dynamic Stocks and Flows (DSF) (please see: http://www.hss.cmu.edu/departments/sds/ddmlab/modeldsf/results.html for details). The DSF task was designed to be as simple and accessible as possible to computational modelers while focusing on two key ubiquitous components of general intelligence: the control of dynamical systems and the prediction of future events. ?A general call for participation was submitted to invite independent modelers using distinct computational approaches to simulate human performance in DSF. ?Participants in this challenge developed computational models to simulate human performance on the DSF task in a variety of conditions. ?The goal was to reproduce human behavior, including learning, mistakes and limitations in a way that would generalize to new conditions of the task undisclosed to the modelers. ?Results from three of the models submitted were selected for presentation at the 2009 International Conference on Cognitive Modeling. ?Human learning data in DSF as well as the results from all the models participating in the model comparison are available on the comparison web site and can be used for the purposes of analyses and publication in this special issue. ? We welcome submissions from those who participated in the DSF model comparison challenge as well as from those who are in a position to comment on the following general topics relevant to model comparison within the context of the DSF challenge: ? ? ???????Many computational fields have seen the emergence of challenge tasks to prod the development of new techniques and measure their progress toward the goal (e.g., Robocup). ?What are the requirements of such challenge tasks for AGI? ?Should they provide independent tests of specific capacities, integrated tests of functionality, or both? ? ???????Progress is often measured on the relative evaluation of alternatives in a common setting. ?But what are the constraints of such comparisons for cognitive models? ?Are acceptable mechanisms limited to those that are judged cognitively - or even biologically - plausible? Should the complexity of a model be taken into account? ?Which levels of description are acceptable? Should models aim to predict human performance in new conditions, or is suitable post hoc reproduction of known performance data sufficient? ? ???????The methodology developed by cognitive psychology for evaluating fits of model to human data is strongly dependent upon experimental control and scales poorly to complex, open-ended tasks. Sets of criteria for evaluating cognitive architectures have been proposed, but specific instantiations on AGI-level tasks have been lacking. ? ???????Human behavior models based on cognitive architectures are usually developed for very specific tasks and at substantial effort to the modeler. ?While cognitive architectures keep being refined, cumulative progress in the form of model reuse has been elusive. New mechanisms and/or practices for composing and/or generalizing models of simple tasks are required for scaling up to models suitable for general, open-ended intelligence. ? ???????Despite their stated goal of providing an integrated theory of human intelligence, specific cognitive architectures are usually applied to a relatively narrow set of cognitive activities, often laboratory tasks. ?Attempts to apply cognitive architectures to open-ended, naturalistic environments (using virtual or robotic embodiments) have raised substantial issues about their robustness and scalability beyond laboratory environments. Submission Submissions should be sent by December 1st, 2009 to DSFChallenge at gmail.com. Manuscripts should conform to the JAGI formatting guidelines that can be found at the journal?s web site, http://journal.agi-network.org/, and should not exceed 20 pages of total length. ?Manuscripts will be submitted to a traditional anonymous peer-review process with publication of accepted contributions expected by summer 2010. ?Authors will be required to provide the final camera-ready, formatted and copy-edited manuscript. Inquiries regarding this special issue can be sent to DSFChallenge at gmail.com or directly to any of the special issue editors at the addresses below. Special Issue Editors Christian Lebiere Psychology Department, Carnegie Mellon University cl at cmu.edu Cleotilde Gonzalez Social and Decision Sciences Department, Carnegie Mellon University coty at cmu.edu Walter Warwick MA&D Operation, Alion Science and Technology wwarwick at alionscience.com -------------- next part -------------- An HTML attachment was scrubbed... URL: From D.H.van.Rijn at rug.nl Wed Sep 16 10:22:12 2009 From: D.H.van.Rijn at rug.nl (Hedderik van Rijn) Date: Wed, 16 Sep 2009 16:22:12 +0200 Subject: [ACT-R-users] Two PhD positions in Groningen Message-ID: Two PhD positions in Groningen: * Cognitive models as believable game opponents (advertised earlier, but please note the chance of deadline!) * The Role of Attention and Memory in Temporal Performance Please see below for the extended descriptions. 4-year PhD position in Experimental Psychology/Cognitive Modeling - University of Groningen The Role of Attention and Memory in Temporal Performance Although no specific sense has been identified that registers time, humans, as most animals, have a very accurate system that governs the perception of time. However, attention strongly influences time perception such that if attention is diverted from the temporal task during the production of a previously learned interval, estimates of that temporal interval are significantly lengthened and more variable. This phenomenon is often used in applied settings to assess mental workload: If estimated times during work become longer and more variable, it is assumed that mental workload is increased. The goal of this project is to explain these findings in terms of the theory of temporal cognition proposed in Taatgen, Van Rijn & Anderson (2007), Van Rijn & Taatgen (2008), and Meijering & Van Rijn (2009). According to this theory, attention or mental workload does not influence time perception itself, but rather affects the more general aspects of cognitive performance. The project will consist of experimental work with human subjects to gain insight in the role of attention and memory processes in temporal cognition, using behavioral studies (e.g., eye-tracking studies), ERP and/or TMS studies, and possibly cognitive modeling using the ACT-R cognitive architecture. Earlier work on time perception will serve as a starting point of the project. Environment The project will be supervised by dr. Hedderik van Rijn (http://www.van-rijn.org/ ) and prof. dr. Addie Johnson, and will be carried out in the Human Performance & Ergonomics group of the Department of Psychology of the University of Groningen. The HP&E group is an internationally oriented research group with multiple researchers working on topics related to attention, memory and task performance. Close collaboration is expected with the Cognitive Modeling group (http://research.ai.rug.nl/index.php/cogmod ) of the Artificial Intelligence department. A monthly stipend will be provided of approximately 1400 Euro. The four-year Ph.D project will be carried out within the Graduate School of the Faculty of Behavioural and Social Sciences and research school of Behavioural and Cognitive Neuroscience. Assistance with residence permits in the Netherlands for non-EU citizens will be provided by the university. Requirements We are looking for a non-Dutch candidate with a degree in Cognitive Science, Experimental/Cognitive (Neuro)Psychology or Artificial Intelligence, with a keen interest in model-based explanations of human cognitive performance. Candidates should have experience with at least two of the following domains: analyzing complex data, programming, or ERPs. In addition, it is advantageous to have experience with cognitive modeling. The candidate must be able to start no later than 15 December, 2009. Applications Send your application, together with a curriculum vitae and the names of two references as (preferably PDF) attachment by e-mail to d.h.van.rijn at rug.nl Deadline for applications is October 4th. On October 4th, the received applications will be examined, but applications arriving after that date may still be considered, until a suitable candidate has been found. We are aiming for a quick selection process. Information about the scholarship, work content etc. can be obtained by writing to dr. D. H. van Rijn, +31 50 363 6290 or email d.h.van.rijn at rug.nl . -------------------------------------- 4 year PhD position in Cognitive Modeling - University of Groningen Cognitive models as believable game opponents The goal of the project is to create cognitive models of how people play games, and use such models to create computer opponents that behave like human opponents. The main challenge of computer (AI) opponents in games has most often been to create an opponent that is smart enough to be able to compete with human players. However, as the classical case of chess illustrates, this does not necessarily mean that the computer plays like a human. Creating an opponent that plays like a human has the advantage that it creates a much richer game experience, especially if the computer player is capable of learning, and can adapt its expertise to the expertise of the player. A pilot project with the game of Set (http://www.ai.rug.nl/~niels/set-app/index.html ) has already demonstrated these virtues. From a scientific point of view, studying human cognition in the context of games offers broad opportunities to gain insights in how players reason about games and their opponents. The project will consist of experimental work with human subjects to gain insight in human performance and learning in games, using eye- tracking to collect fine-grained data, cognitive modeling using the ACT-R cognitive architecture, and construction of computer game opponents. Earlier work on the Set game will serve as a starting point of the project, and will be extended to other games as the project progresses. Environment The project will be supervised by Prof. dr. Niels Taatgen (http://www.ai.rug.nl/~niels/ ), and will be carried out in the Cognitive Modeling group (http://research.ai.rug.nl/index.php/cogmod ) of the Artificial Intelligence department of the University of Groningen. A monthly stipend will be provided of approximately 1400 Euro. The four-year Ph.D project will be carried out within the school of behavioral and cognitive neuroscience. Assistance with stay permits in the Netherlands for non- EU citizens will be provided by the university. Requirements We are looking for a candidate with a degree in Cognitive Science, Artificial Intelligence, Cognitive Psychology or Computer Science with experience in programming and a keen interest in psychological processes. In addition, it is advantageous to have experience with cognitive modeling. Applications Write your application, together with a curriculum vitae and the names of two references as (preferably PDF) attachment by e-mail to .a.taatgen at rug.nl On October 4th, the received applications will be examined, but applications arriving after that date may still be considered, until a suitable candidate has been found. Information about the scholarship, work content etc may be provided by Prof.dr. N.A. Taatgen, +31 50 3636435 or email n.a.taatgen at rug.nl). From L.van.Maanen at ai.rug.nl Fri Sep 18 03:35:11 2009 From: L.van.Maanen at ai.rug.nl (Leendert van Maanen) Date: Fri, 18 Sep 2009 09:35:11 +0200 Subject: [ACT-R-users] Need for Multi-Level Activation Spread in ACT-R In-Reply-To: References: Message-ID: Jerry, When developing a new, more detailed retrieval mechanism for ACT-R (called RACE/A, see also www.ai.rug.nl/~leendert/race), we ran into similar issues. That is, we were interested in multilink spreading activation to compute the activation levels of multiple chunks *during* a memory retrieval. In addition, for our model of picture- word interference (Van Maanen & Van Rijn, 2007 CSR) we also required the capability of spreading activation from a stimulus (a word or a picture) to multiple chunks, analoguous in your airspeed/arispeed example. To achieve this, we used the retrieval-set-hook to compute activations and add-sji to manually set the spreading activation between the chunks. This is similar to what Dan suggested. One of the virtues of our approach is that during retrieval of one chunk other, associated (set with add-sji), chunks will also increase in activation, allowing for the interactions between pos/letter information that you are interested in. One of the drawbacks is that computations will slow down tremendously, as you said. We haven't tested RACE/A on large DMs, but with a recalculation of every activation value every 5ms model runs take long, even for small DM sizes. Leendert On 14Sep 2009, at 19:45 , Ball, Jerry T Civ USAF AFMC 711 HPW/RHAC wrote: > We are in the process of mapping the linguistics representations > that are generated by our language comprehension model into a > situation model based semantic representation. We are trying to do > this in a representationally reasonable way within the ACT-R > architecture. The problem we face is the many-to-many mapping > between words and concepts. Individual words may map to multiple > concepts, and individual concepts may may to multiple words. Given > this many-to-many mapping, we would like to use mapping chunks to > map from words to concepts. The mapping chunks would encode a single > mapping relationship (e.g. a separate mapping chunk to map from the > word "bank" to the financial institution concept; from the word > "bank" to the river bank concept; from the concept dog to the word > "dog"; from the concept dog to the word "canine"). When processing a > word, the goal is to retrieve the contextually relevant concept. We > would like to accomplish this in a single retrieval, however, we do > not know how to do this given the single-level activation spreading > mechanism in ACT-R. Since there is no direct link between a word and > a concept if mapping chunks are used (i.e. there is no slot in the > concept that contains the word), the word will not spread activation > to the concept. Instead, given the use of mapping chunks, it appears > that two retrievals are needed: 1) given the word, retrieve a > mapping chunk, and 2) given a mapping chunk, retrieve a concept. > Since our model of language comprehension is already slower than > humans at processing language, any extra retrievals are problematic. > In fact, we have already eliminated an extra retrieval in > determining the part-of-speech of a word. Previously, two retrievals > were needed: 1) retrieve the word corresponding to the perceptual > input, and 2) given the word (and context) retrieve the part-of- > speech of the word. While we were successful in eliminating a > retrieval, the resulting word-pos chunks contain a mixture of word > form information (e.g. the letters and trigrams in the word) and pos > information. Even so, they do not yet contain any representation of > phonetic, phonemic, syllabic or morphemic information. With just > letter and trigram information, long words contain many slots. > Ideally, we would like to represent letter and trigram information > independently of each other and POS information (allowing them to > interact in retrieving a word), but given the single-level > activation spreading mechanism in ACT-R doing so would necessitate > multiple independent retrievals, which would fail to capture the > interaction of letter and trigram information that leads to > successful retrievals of words in the face of variability in the > perceptual form (e.g. "arispeed" should retrieve "airspeed"). > > The fall back for mapping words to concepts is to embed all > the possible concepts as slot values in a word and vice versa. While > we consider this a representationally problematic solution -- word > and concept chunks will wind up needing many extra slots, we do not > know how else to work around the single-level activation spread in > ACT-R. > > The primary empirical argument against the need for multi- > level activation spread in ACT-R is based on studies which show no > activation from words like "bull" to words like "milk", even though > "bull" activates "cow" and "cow" activates "milk". Even if it is > true that there are no instances of "indirect" activation from > "bull" to "milk", this does not rule out the need for multi-level > activation spread. There is a hidden assumption that "cow" and > "bull" are directly associated, and that "cow" and "milk" are also > directly associated. Such direct associations may seem reasonable in > small-scale models addressing specific spreading activation > phenomena, but they are questionable in a larger-scale model. Do we > really want to include all the direct associates of "cow" as slot > values in the "cow" chunk, and do the same for all other chunks? > > We understand that the inclusion of a multi-level activation > spreading mechanism in ACT-R would be computationally explosive. > However, we would like to have the capability to explore use of such > a mechanism and to look for ways to keep it computationally > tractable. We have already dealt with the problem of computational > explosion in our word retrieval mechanism. Originally, we attempted > to use a "soft constraint" retrieval mechanism for words. All words > in DM were candidates for retrieval--the most highly activated word > being retrieved. With just 2500 words in DM, the activation > calculations slowed the model down considerably. To manage > retrievals in a tractable manner we implemented a disjunctive > retrieval capability combined with a new perceptual span mechanism > -- the model first tries a hard-constraint retrieval on the entire > perceptual span (which is larger than a word) using the "get-chunk" > function (and chop-string under the covers). If get-chunk succeeds > (indicating that there is a chunk in DM corresponding to the entire > perceptual span) a retrieval is constructed using the entire > perceptual span as a hard constraint to retrieve the corresponding > multi-word unit in DM, if this fails, the model backs-off and uses > the first space delimited word (using chop-string) in the perceptual > span to check for a corresponding word in DM -- if a match is found > with get-chunk, a retrieval is constructed to retrieve the word. If > all else fails, we construct a retrieval that imposes a hard > constraint on the first letter (this is less than ideal, but a > reasonable compromise). The overall effect is a (nearly) soft- > constraint retrieval implemented in a computationally tractable way. > > A similar capability to effect multi-level activation spread > in a computationally tractable manner would be highly desirable. > > Jerry > > _______________________________________________ > 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 ########################### Leendert van Maanen Department of Artificial Intelligence University of Groningen P.O.Box 407 9700 AK Groningen The Netherlands W: http://www.ai.rug.nl/~leendert E: leendert at ai.rug.nl T: +31 50 363 7603 ########################### -------------- next part -------------- An HTML attachment was scrubbed... URL: From byrne at rice.edu Tue Sep 22 17:28:52 2009 From: byrne at rice.edu (Mike Byrne) Date: Tue, 22 Sep 2009 16:28:52 -0500 Subject: [ACT-R-users] Postdoc position: Cognitive modeling and aviation human factors Message-ID: Postdoctoral Research Position in Cognitive Modeling and Aviation Human Factors The Computer-Human Interaction Laboratory at Rice University currently has an opening on a NASA-sponsored project entitled "A Transparent Research Environment for Aviation Modeling." This work is focussed on ACT-R computational modeling of commercial airline pilots with the larger goal of informing large-scale simulations of air traffic management systems as part of the next generation of U.S. air traffic management. The primary responsibilities of the postdoctoral researcher will be performing task analyses of benchmark piloting tasks, transforming those task analyses into ACT-R models, and interfacing those models with a medium-fidelity flight simulation environment, X-Plane. Collaborators at the University of Illinois will provide human data for model comparison and validation. Requirements include a Ph.D. in cognitive science, computer science or psychology with cognitive modeling experience, preferably but not necessarily using the ACT-R cognitive architecture. Previous experience with aviation human factors is also desirable but not necessary. U.S. citizenship or permanent residency (i.e., Green Card) is required. Applications will be reviewed beginning on October 1, 2009 continuing until the position is filled. The position may be available as soon as October 15, 2009. Salary is expected to start at approximately $45,000 per year depending on qualifications. Interested individuals should send a curriculum vitae, a statement of research interests, names of three references, and a cover letter to Mike Byrne at byrne at rice.edu or via snail mail: Mike Byrne Rice University Department of Psychology 6100 Main St., MS-25 Houston, TX 77005 From tiffany.Jastrzembski at mesa.afmc.af.mil Sat Sep 26 11:30:27 2009 From: tiffany.Jastrzembski at mesa.afmc.af.mil (Jastrzembski, Tiffany S Civ USAF AFMC 711 HPW/RHAC) Date: Sat, 26 Sep 2009 11:30:27 -0400 Subject: [ACT-R-users] Call for Papers: Behavioral Representation in Modeling & Simulation (BRIMS) 2010 In-Reply-To: References: Message-ID: <2B00361EE3107A4F88383EC1B041DC9A05FB94FF@VFOHMLAO01.Enterprise.afmc.ds.af.mil> You are invited to participate in the 19th Conference on Behavior Representation in Modeling and Simulation (BRIMS). BRIMS enables modeling and simulation research scientists, engineers, and technical communities across disciplines to meet, share ideas, identify capability gaps, discuss cutting-edge research directions, highlight promising technologies, and showcase the state-of-the-art in applications. The BRIMS Conference will consist of many exciting elements in 2010, including special topic areas, technical paper sessions, special symposia/panel discussions, and government laboratory sponsor sessions. For additional information, the BRIMS 2010 official website will open early next week (www.brimsconference.org). BRIMS 2010 includes a dynamic and eclectic lineup of keynote speakers, including: Wayne Gray, PhD Rensselaer Polytechnic Institute, http://www.rpi.edu/~grayw/ LCDR Joseph Cohn, Phd DARPA, http://www.darpa.mil/dso/personnel/cohn.htm Jerrold Post, MD George Washington University, http://www.gwu.edu/~elliott/faculty/post.cfm Robert Axtell, PhD George Mason University, http://www.santafe.edu/profiles/?pid=79 The BRIMS Executive Committee invites papers, posters, demos, symposia, panel discussions, and tutorials on topics related to the representation of individuals, groups, teams and organizations in models and simulations. All submissions are peer-reviewed. Key Dates: All submissions due: December 21, 2009 Tutorial Acceptance: February 1, 2010 Authors Notification February 1, 2010 Final version due: February 19, 2010 Tutorials held: March 22, 2010 BRIMS 2010 Opens: March 23, 2010 Special Topic Areas of Interest The following research arenas are identified to elicit specific technical content: * Socio-cultural modeling and simulation * Neurobiological & biologically-inspired cognitive modeling * Bridging the M&S gap between the laboratory and field settings * Current and future M&S challenges and solutions * Model validation * Necessity & sufficiency of mechanisms and parameters * Model comparison General Topic Areas of Interest General areas of interest include, but are not limited to: Modeling * Cognitive or behavioral moderators on performance * Intelligent agents and avatars * Models of reasoning and decision making * Team, group, crowd, and organizational behavior * Physical models of human movement * Performance assessment and skill monitoring/tracking * Performance prediction * Performance enhancement/optimization * Modeling architectures/knowledge representation systems * Knowledge acquisition/engineering * Human behavior issues in model federations * Human behavior representation for system design and evaluation Simulation * Synthetic environments for human behavior representation * Terrain representation and reasoning * Spatial reasoning * Time representation * Human behavior usability and interoperability * Efficiency, usability, affordability issues * Operator interfaces * Multi-resolution/fidelity simulations Types of Submissions Paper Presentations Paper Presentation Sessions are composed of 3 to 4 presentations on a related topic. Presentations are done lecture-style, allowing 20 minutes for presentation content and 5 minutes for questions. Papers should describe original research that has not been published elsewhere. Accepted papers are published in the Proceedings. Papers not accepted as full papers will be considered for poster presentations. Paper presentations must be submitted as full papers, ranging between 5 to 8 pages in length. Symposia/Panel Discussions These sessions are 60-90 minutes long, and encompass several speakers presenting research and/or engaging in discussion on related aspects of a common topic of interest to the BRIMS community. These are not merely collections of presentations; rather, they should consist of a set of common questions/issues addressed by all participants. Abstracts for selected symposia/panel discussions will be published in the Proceedings. Submissions for symposia or panel discussions must consist of a 2 to 4 page abstract with a session title, identification of chair, brief statements (approximately 250 words) from each participant summarizing main focal points, and identification of common questions/issues addressed by all discussants. Interactive Session: Posters and Demos The Interactive Session involves a longer (approximately 2 hour) period of multiple simultaneous presentations and provides an opportunity for continuous interaction with conference attendees. This session features both static posterboard displays and live demonstrations of state-of-the-art research or technologies. Accepted abstracts are published in the Proceedings. Poster submissions are limited to a 2-page extended abstract describing the research to be presented or the application/technology to be demonstrated. The Interactive Session will be held the opening evening, March 23, 2010. Interactive Session: Exhibitor Session The Exhibitor Session creates a forum to display and demonstrate leading technology and applications and will co-occur with the poster session. This is an opportunity for the BRIMS community to see firsthand, the latest advances and capabilities in modeling and simulation tools. Exhibitor submissions are limited to a one page description. Please contact Jeanne Eury (jeury at lodestar-group.com / Phone 919-326-0278) for exhibitor fees. Tutorials Tutorials provide conference participants the opportunity to gain new insights, knowledge, and skills in an area related to the interests of the BRIMS community. Tutorials are presented in a lecture-and-discussion or learning-by-doing format. Tutorials may be a half-day (3 hours, plus breaks) or a full-day (6 hours, plus breaks) in duration, and will take place on Monday, March 22, 2010. Tutorial proposals may be submitted through the on-line submission system and descriptions for accepted tutorials will be included in conference announcements and in the Proceedings. Tutorial descriptions should include a detailed outline of the material that will be covered. Submission Process and Format Submissions are handled on-line at the BRIMS website, visit www.brimsconference.org for online submissions. Please see the guidelines on the BRIMS website for format requirements and content suggestions. If you have any questions about the submission process or are unable to submit to the web site, please contact Jeanne Eury by email (jeury at lodestar-group.com) or phone 919-326-0278. ACCOMMODATIONS and REGISTRATION The conference will be held at the Charleston Harbor Resort & Marina with conference rates and limited government per diem rates available. Visit www.charlestonharborresort.com for general information about the site and accommodations. Conference and hotel registration, general area, and travel information can be found at www.brimsconference.org. ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Tiffany S. Jastrzembski, Ph.D. BRIMS 2010 Conference Chair Cognitive Research Psychologist 711th Human Performance Wing Air Force Research Laboratory 6030 South Kent Street, Mesa, AZ 85212 Phone: (480) 988-6561 x688 tiffany.jastrzembski at mesa.afmc.af.mil From no-reply at getdropbox.com Tue Sep 29 23:16:04 2009 From: no-reply at getdropbox.com (Dropbox) Date: Wed, 30 Sep 2009 03:16:04 +0000 Subject: [ACT-R-users] Jibo He has invited you to Dropbox Message-ID: <20090930031604.0BC025DE6AC@web6.getdropbox.com> We're excited to let you know that Jibo He has invited you to Dropbox! Jibo He has been using Dropbox to sync and share files online and across computers, and thought you might want it too. Visit http://www.getdropbox.com/link/20.l1IJ4lRTFp/NjMyMTUxNzA3 to get started. - The Dropbox Team ____________________________________________________ To stop receiving invites from Dropbox, please go to http://www.getdropbox.com/bl/bc77674b194e/act-r-users%40act-r.psy.cmu.edu -------------- next part -------------- An HTML attachment was scrubbed... URL: