DOCSLIB.ORG

Connectionist Models of Language Processing

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Connectionist Models of Language Processing Cognitive Studies Preprint 2003, Vol. 10, No. 1, 10–28 Connectionist Models of Language Processing Douglas L. T. Rohde David C. Plaut Massachusetts Institute of Technology Carnegie Mellon University Traditional approaches to language processing have been based on explicit, discrete represen- tations which are difficult to learn from a reasonable linguistic environment—hence, it has come to be accepted that much of our linguistic representations and knowledge is innate. With its focus on learning based upon graded, malleable, distributed representations, connectionist modeling has reopened the question of what could be learned from the environment in the ab- sence of detailed innate knowledge. This paper provides an overview of connectionist models of language processing, at both the lexical and sentence levels. Although connectionist models have been applied to the full and processes of actual language users: Language is as lan- range of perceptual, cognitive, and motor domains (see Mc- guage does. In this regard, errors in performance (e.g., “slips Clelland, Rumelhart, & PDP Research Group, 1986; Quin- of the tongue”; Dell, Schwartz, Martin, Saffran, & Gagnon, lan, 1991; McLeod, Plunkett, & Rolls, 1998), it is in their 1997) are no less valid than skilled language use as a measure application to language that they have evoked the most in- of the underlying nature of language processing. The goal is terest and controversy (e.g., Pinker & Mehler, 1988). This not to abstract away from performance but to articulate com- is perhaps not surprising in light of the special role that lan- putational principles that account for it. guage plays in human cognition and culture. It also stems in A major attraction of the connectionist approach to lan- part from the considerable difference in goals and methods guage, apart from its natural relation to neural computation, between linguistic and psychological approaches to the study is that the very same processing mechanisms apply across of language. This rift goes deeper than a simple dichotomy the full range of linguistic structure. This paper provides an of emphasizing competence versus performance (Chomsky, overview of connectionist models of language processing, at 1957)—it cuts to the heart of the question of what it means both the lexical and sentence levels. to know and use a language (Seidenberg, 1997). Traditional approaches to language processing have been Lexical Processing based on explicit, discrete representations which are difficult or impossible to learn from a reasonable linguistic environ- Phonological development ment (Gold, 1967). Therefore, it has come to be accepted that much of our linguistic representations and knowledge Although the use of language seems straightforward to is innate. With its focus on learning based upon graded, adult native speakers, an infant must solve numerous diffi- malleable, distributed representations, connectionist model- cult computational problems in learning to understand and ing has reopened the question of what could be learned from produce speech, stemming from the fact that speech is ex- the environment in the absence of detailed innate knowledge. tended in time, highly variable and, at a morphemic level, Although the need to learn internal representations poten- has no systematic relation to its underlying meaning. More- tially gives connectionist networks great power and flexibil- over, infants must learn to produce comprehensible speech ity, it also introduces limitations. These limitations are im- without any direct articulatory instruction or feedback. portant and, ideally, will reflect limitations observed in hu- Plaut and Kello (1999) proposed a framework for phono- man language processing. logical development in which phonology mediates among acoustic, articulatory, and semantic representations in the From a connectionist perspective, performance is not an service of both comprehension and production. A critical imperfect reflection of some abstract competence, but rather aspect of the approach is that, given the absence of direct ar- the behavioral manifestation of the internal representations ticulatory feedback, learning to produce speech is driven by indirect feedback derived from the comprehension system— that is, from the acoustic, phonological, and semantic con- sequences of the system’s own articulations (Locke, 1983; This research was supported by an NIH Integrated Behavioral Science Center Grant (MH64445, J. McClelland, PD), and by Menn & Stoel-Gammon, 1995; Studdert-Kennedy, 1993). NIMH NRSA MH65105. Correspondence regarding this article This is accomplished by learning an internal forward model may be sent either to Douglas Rohde ([email protected]), MIT, of the physical processes that relate articulation to acoustics NE20-437E, 3 Cambridge Center, Cambridge, MA 02139, USA, (Jordan & Rumelhart, 1992). Such a model is learned by or to David Plaut ([email protected]), Department of Psychology, executing a variety of articulations, predicting how they will Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, PA sound, and then adapting the model based on the discrepancy 15213-3890, USA. between this prediction and the actual resulting acoustics. In 1 2 ROHDE AND PLAUT the infant, the forward model is assumed to develop primarily rich languages like Hebrew (e.g., Frost, Deutsch, & Forster, as a result of reduplicated and variegated babbling in the sec- 2000) are typically interpreted as being problematic for the ond half of the first year (Vihman, 1996). Once developed, connectionist account. To evaluate whether this interpreta- the forward model can be used to convert acoustic feedback tion is valid, Plaut and Gonnerman (2000) carried out simu- (i.e, whether an utterance sounded right) into the articula- lations in which a set of morphologically related words vary- tory feedback necessary to train speech production (Perkell, ing in semantic transparency were embedded in either a mor- Matthies, Svirsky, & Jordan, 1995). An implementation of phologically rich or impoverished artificial language. They the framework, in the form of a simple recurrent network found that morphological priming increased with degree of (Elman, 1991a), learned to comprehend, imitate, and inten- semantic transparency in both languages. Critically, priming tionally name a corpus of 400 monosyllabic words, and its extended to semantically opaque items in the morphologi- speech errors in development were similar to those of young cally rich language (consistent with findings in Hebrew) but children. not in the impoverished language (consistent with findings in English). Such priming arises because the processing of all Morphology items, including opaque forms, is influenced by the degree of morphological organization of the entire system. These Most linguistic domains are quasi-regular in that there is findings suggest that, rather than being challenged by the considerable systematicity between inputs and outputs but occurrence of non-semantic morphological effects in mor- also numerous exceptions. A standard assumption is that phologically rich languages, the connectionist approach may systematic linguistic knowledge takes the form of explicit provide an explanation for the cross-linguistic differences in rules and that items which violate the rules are handled by a the occurrence of these effects. separate associative mechanism (see Pinker, 1999). Connec- tionist modeling provides an alternative view, in which all Word reading items coexist within a single system whose representations and processing reflect the relative degree of consistency in Many of the issues concerning quasi-regularity in mor- the mappings for different items. phological processing also arise in the context of word read- A key battleground in the debate between these two views ing. As in morphology, the spelling-sound correspondences of the language system has been the relatively constrained of English are highly systematic but admit many excep- domain of English inflectional morphology—specifically, tions (e.g., HAVE, PINT, YACHT) and, as in morphology, re- forming the past-tense of verbs. Rumelhart and Mc- searchers have proposed separate mechanisms for processing Clelland (1986) attempted to reformulate the issue away regular and exception items (Coltheart, Rastle, Perry, Lang- from a sharp dichotomy between explicit rules (add –ed; don, & Ziegler, 2001). e.g., WALK/WALKED) and exceptions (e.g., SING/SANG, Plaut, Seidenberg, McClelland, and Patterson (1996) de- DRINK/DRANK, GO/WENT), and toward a view that empha- veloped a series of connectionist simulations in support of an sizes the graded structure relating verbs and their inflections. alternative conception of language knowledge and process- They developed a connectionist model that learned a direct ing in which all items coexist within a single system whose association between the phonology of all types of verb stems representations and processing reflect the relative degree of and the phonology of their past-tense forms. Although this consistency in the mappings for different items. Different initial model had numerous limitations (Pinker & Prince, types of information about a word—orthographic, phonolog- 1988), many of these have been addressed in subsequent ical, and semantic—are represented as distributed patterns simulation work (Cottrell & Plunkett, 1995; MacWhinney over separate groups of units. In performing a task like read- & Leinbach, 1991; Marchman, 1993; Plunkett & Marchman, ing aloud, orthographic information
Load more

Recommended publications
  • A Distributed Computational Cognitive Model for Object Recognition
    SCIENCE CHINA xx xxxx Vol. xx No. x: 1{15 doi: A Distributed Computational Cognitive Model for Object Recognition Yong-Jin Liu1∗, Qiu-Fang Fu2, Ye Liu2 & Xiaolan Fu2 1Tsinghua National Laboratory for Information Science and Technology, Department of Computer Science and Technology, Tsinghua University, Beijing 100084, China, 2State Key Lab of Brain and Cognitive Science, Institute of Psychology, Chinese Academy of Sciences, Beijing 100101, China Received Month Day, Year; accepted Month Day, Year Abstract Based on cognitive functionalities in human vision processing, we propose a computational cognitive model for object recognition with detailed algorithmic descriptions. The contribution of this paper is of two folds. Firstly, we present a systematic review on psychological and neurophysiological studies, which provide collective evidence for a distributed representation of 3D objects in the human brain. Secondly, we present a computational model which simulates the distributed mechanism of object vision pathway. Experimental results show that the presented computational cognitive model outperforms five representative 3D object recognition algorithms in computer science research. Keywords distributed cognition, computational model, object recognition, human vision system 1 Introduction Object recognition is one of fundamental tasks in computer vision. Many recognition algorithms have been proposed in computer science area (e.g., [1, 2, 3, 4, 5]). While the CPU processing speed can now be reached at 109 Hz, the human brain has a limited speed of 100 Hz at which the neurons process their input. However, compared to state-of-the-art computational algorithms for object recognition, human brain has a distinct advantage in object recognition, i.e., human being can accurately recognize one from unlimited variety of objects within a fraction of a second, even if the object is partially occluded or contaminated by noises.
    [Show full text]
  • The Place of Modeling in Cognitive Science
    The place of modeling in cognitive science James L. McClelland Department of Psychology and Center for Mind, Brain, and Computation Stanford University James L. McClelland Department of Psychology Stanford University Stanford, CA 94305 650-736-4278 (v) / 650-725-5699 (f) [email protected] Running Head: Modeling in cognitive science Keywords: Modeling frameworks, computer simulation, connectionist models, Bayesian approaches, dynamical systems, symbolic models of cognition, hybrid models, cognitive architectures Abstract I consider the role of cognitive modeling in cognitive science. Modeling, and the computers that enable it, are central to the field, but the role of modeling is often misunderstood. Models are not intended to capture fully the processes they attempt to elucidate. Rather, they are explorations of ideas about the nature of cognitive processes. As explorations, simplification is essential – it is only through simplification that we can fully understand the implications of the ideas. This is not to say that simplification has no downsides; it does, and these are discussed. I then consider several contemporary frameworks for cognitive modeling, stressing the idea that each framework is useful in its own particular ways. Increases in computer power (by a factor of about 4 million) since 1958 have enabled new modeling paradigms to emerge, but these also depend on new ways of thinking. Will new paradigms emerge again with the next 1,000-fold increase? 1. Introduction With the inauguration of a new journal for cognitive science, thirty years after the first meeting of the Cognitive Science Society, it seems essential to consider the role of computational modeling in our discipline.
    [Show full text]
  • A Bayesian Computational Cognitive Model ∗
    A Bayesian Computational Cognitive Model ∗ Zhiwei Shi Hong Hu Zhongzhi Shi Key Laboratory of Intelligent Information Processing Institute of Computing Technology, CAS, China Kexueyuan Nanlu #6, Beijing, China 100080 {shizw, huhong, shizz}@ics.ict.ac.cn Abstract problem solving capability and provide meaningful insights into mechanisms of human intelligence. Computational cognitive modeling has recently Besides the two categories of models, recently, there emerged as one of the hottest issues in the AI area. emerges a novel approach, hybrid neural-symbolic system, Both symbolic approaches and connectionist ap- which concerns the use of problem-specific symbolic proaches present their merits and demerits. Al- knowledge within the neurocomputing paradigm [d'Avila though Bayesian method is suggested to incorporate Garcez et al., 2002]. This new approach shows that both advantages of the two kinds of approaches above, modal logics and temporal logics can be effectively repre- there is no feasible Bayesian computational model sented in artificial neural networks [d'Avila Garcez and concerning the entire cognitive process by now. In Lamb, 2004]. this paper, we propose a variation of traditional Different from the hybrid approach above, we adopt Bayesian network, namely Globally Connected and Bayesian network [Pearl, 1988] to integrate the merits of Locally Autonomic Bayesian Network (GCLABN), rule-based systems and neural networks, due to its virtues as to formally describe a plausible cognitive model. follows. On the one hand, by encoding concepts
    [Show full text]
  • Cognitive Psychology for Deep Neural Networks: a Shape Bias Case Study
    Cognitive Psychology for Deep Neural Networks: A Shape Bias Case Study Samuel Ritter * 1 David G.T. Barrett * 1 Adam Santoro 1 Matt M. Botvinick 1 Abstract to think of these models as black boxes, and to question whether they can be understood at all (Bornstein, 2016; Deep neural networks (DNNs) have advanced Lipton, 2016). This opacity obstructs both basic research performance on a wide range of complex tasks, seeking to improve these models, and applications of these rapidly outpacing our understanding of the na- models to real world problems (Caruana et al., 2015). ture of their solutions. While past work sought to advance our understanding of these models, Recent pushes have aimed to better understand DNNs: none has made use of the rich history of problem tailor-made loss functions and architectures produce more descriptions, theories, and experimental methods interpretable features (Higgins et al., 2016; Raposo et al., developed by cognitive psychologists to study 2017) while output-behavior analyses unveil previously the human mind. To explore the potential value opaque operations of these networks (Karpathy et al., of these tools, we chose a well-established analy- 2015). Parallel to this work, neuroscience-inspired meth- sis from developmental psychology that explains ods such as activation visualization (Li et al., 2015), abla- how children learn word labels for objects, and tion analysis (Zeiler & Fergus, 2014) and activation maxi- applied that analysis to DNNs. Using datasets mization (Yosinski et al., 2015) have also been applied. of stimuli inspired by the original cognitive psy- Altogether, this line of research developed a set of promis- chology experiments, we find that state-of-the-art ing tools for understanding DNNs, each paper producing one shot learning models trained on ImageNet a glimmer of insight.
    [Show full text]
  • Cognitive Modeling, Symbolic
    Cognitive modeling, symbolic Lewis, R.L. (1999). Cognitive modeling, symbolic. In Wilson, R. and Keil, F. (eds.), The MIT Encyclopedia of the Cognitive Sciences. Cambridge, MA: MIT Press. Symbolic cognitive models are theories of human cognition that take the form of working computer programs. A cognitive model is intended to be an explanation of how some aspect of cognition is accomplished by a set of primitive computational processes. A model performs a specific cognitive task or class of tasks and produces behavior that constitutes a set of predictions that can be compared to data from human performance. Task domains that have received considerable attention include problem solving, language comprehension, memory tasks, and human-device interaction. The scientific questions cognitive modeling seeks to answer belong to cognitive psychology, and the computational techniques are often drawn from artificial intelligence. Cognitive modeling differs from other forms of theorizing in psychology in its focus on functionality and computational completeness. Cognitive modeling produces both a theory of human behavior on a task and a computational artifact that performs the task. The theoretical foundation of cognitive modeling is the idea that cognition is a kind of COMPUTATION (see also COMPUTATIONAL THEORY OF MIND). The claim is that what the mind does, in part, is perform cognitive tasks by computing. (The claim is not that the computer is a metaphor for the mind, or that the architectures of modern digital computers can give us insights into human mental architecture.) If this is the case, then it must be possible to explain cognition as a dynamic unfolding of computational processes.
    [Show full text]
  • Connectionist Models of Cognition Michael S. C. Thomas and James L
    Connectionist models of cognition Michael S. C. Thomas and James L. McClelland 1 1. Introduction In this chapter, we review computer models of cognition that have focused on the use of neural networks. These architectures were inspired by research into how computation works in the brain and subsequent work has produced models of cognition with a distinctive flavor. Processing is characterized by patterns of activation across simple processing units connected together into complex networks. Knowledge is stored in the strength of the connections between units. It is for this reason that this approach to understanding cognition has gained the name of connectionism. 2. Background Over the last twenty years, connectionist modeling has formed an influential approach to the computational study of cognition. It is distinguished by its appeal to principles of neural computation to inspire the primitives that are included in its cognitive level models. Also known as artificial neural network (ANN) or parallel distributed processing (PDP) models, connectionism has been applied to a diverse range of cognitive abilities, including models of memory, attention, perception, action, language, concept formation, and reasoning (see, e.g., Houghton, 2005). While many of these models seek to capture adult function, connectionism places an emphasis on learning internal representations. This has led to an increasing focus on developmental phenomena and the origins of knowledge. Although, at its heart, connectionism comprises a set of computational formalisms,
    [Show full text]
  • Artificial Intelligence and Cognitive Science Have the Same Problem
    Artificial Intelligence and Cognitive Science have the Same Problem Nicholas L. Cassimatis Department of Cognitive Science, Rensselaer Polytechnic Institute 108 Carnegie, 110 8th St. Troy, NY 12180 [email protected] Abstract other features of human-level cognition. Therefore, not Cognitive scientists attempting to explain human understanding how the relatively simple and “unintelligent” intelligence share a puzzle with artificial intelligence mechanisms of atoms and molecules combine to create researchers aiming to create computers that exhibit human- intelligent behavior is a major challenge for a naturalistic level intelligence: how can a system composed of relatively world view (upon which much cognitive science is based). unintelligent parts (such as neurons or transistors) behave Perhaps it is the last major challenge. Surmounting the intelligently? I argue that although cognitive science has human-level intelligence problem also has enormous made significant progress towards many of its goals, that technological benefits which are obvious enough. solving the puzzle of intelligence requires special standards and methods in addition to those already employed in cognitive science. To promote such research, I suggest The State of the Science creating a subfield within cognitive science called For these reasons, understanding how the human brain intelligence science and propose some guidelines for embodies a solution to the human-level intelligence research addressing the intelligence puzzle. problem is an important goal of cognitive science. At least at first glance, we are nowhere near achieving this goal. There are no cognitive models that can, for example, fully The Intelligence Problem understand language or solve problems that are simple for a Cognitive scientists attempting to fully understand human young child.
    [Show full text]
  • Models of Cognition: Neurological Possiblity Does Not Indicate Neurological Plausibility
    UC Merced Proceedings of the Annual Meeting of the Cognitive Science Society Title Models of Cognition: Neurological Possiblity Does Not Indicate Neurological Plausibility Permalink https://escholarship.org/uc/item/26t9426g Journal Proceedings of the Annual Meeting of the Cognitive Science Society, 27(27) ISSN 1069-7977 Authors Kriete, Trenton E. Noelle, David C. Publication Date 2005 Peer reviewed eScholarship.org Powered by the California Digital Library University of California Models of Cognition: Neurological possibility does not indicate neurological plausibility. Peter R. Krebs ([email protected]) Cognitive Science Program School of History & Philosophy of Science The University of New South Wales Sydney, NSW 2052, Australia Abstract level cognitive functions like the detection of syntactic and semantic features for words (Elman, 1990, 1993), Many activities in Cognitive Science involve complex learning the past tense of English verbs (Rumelhart and computer models and simulations of both theoretical and real entities. Artificial Intelligence and the study McClelland, 1996), or cognitive development (McLeod of artificial neural nets in particular, are seen as ma- et al., 1998; Schultz, 2003). SRNs have even been jor contributors in the quest for understanding the hu- suggested as a suitable platform \toward a cognitive man mind. Computational models serve as objects of neurobiology of the moral virtues" (Churchland, 1998). experimentation, and results from these virtual experi- ments are tacitly included in the framework of empiri- While some of the models go back a decade or more, cal science. Cognitive functions, like learning to speak, there is still great interest in some of these `classics', or discovering syntactical structures in language, have and similar models are still being developed, e.g.
    [Show full text]
  • Artificial Neural Nets
    Artificial Neural Nets: a critical analysis of their effectiveness as empirical technique for cognitive modelling Peter R. Krebs A thesis submitted in fulfillment of the requirements for the degree of Doctor of Philosophy. University of New South Wales August 2007 COPYRIGHT, AUTHENTICITY, and ORIGINALITY STATEMENT I hereby grant the University of New South Wales or its agents the right to archive and to make available my thesis or dissertation in whole or part in the University libraries in all forms of media, now or here after known, subject to the provisions of the Copyright Act 1968. I retain all proprietary rights, such as patent rights. I also retain the right to use in future works (such as articles or books) all or part of this thesis or dissertation. I also authorise University Microfilms to use the 350 word abstract of my thesis in Dissertation Abstract International. I have either used no substantial portions of copyright material in my thesis or I have obtained permission to use copyright material; where permission has not been granted I have applied/will apply for a partial restric- tion of the digital copy of my thesis or dissertation, and I certify that the Library deposit digital copy is a direct equivalent of the final officially approved version of my thesis. No emendation of content has occurred and if there are any minor variations in formatting, they are the result of the conversion to digital format, and I hereby declare that this submission is my own work and to the best of my knowledge it contains no materials previously published or written by another person, or substantial proportions of material which have been accepted for the award of any other degree or diploma at UNSW or any other educational institution, except where due acknowledgment is made in the thesis.
    [Show full text]
  • A Cognitive Model of Drivers Attention
    A Cognitive Model of Drivers Attention Kerstin Sophie Haring ([email protected]) Research Center for Advanced Science and Technology, The University of Tokyo 4-6-1, Komaba, Meguro-ku, Tokyo, 153-8904, Japan Marco Ragni ([email protected]) Lars Konieczny ([email protected]) Center for Cognitive Science, University of Freiburg Friedrichstr. 50, 79098 Freiburg, Germany Abstract open source LISP, which not only guarantees support and accountability, but also enables the research community to Cognitive architectures can account for highly complex tasks. One of the greatest challenges is understanding and modeling use the developed model for further research. It is able to human driving behavior. This paper describes an integrated keep a traffic lane, initiate and decide about a change of the cognitive model of human attention during the performance lane in case of upfront traffic, identify prevalent situations at of car driving. In this task, the attention process can be crossroads and react to traffic signs. divided into at least three basic components: the control process, the monitoring process, and finally, the decision making process. Of these basic tasks, the first has the highest priority. All three phases are implemented in a cognitive model in the cognitive Architecture ACT-R 6.0. The model is able to keep a traffic lane, overtake another vehicle by lane change, identifies traffic signs and different situations emerging at crossroads. Keywords: Driver behavior model; cognitive architecture; ACT-R; Attention Introduction Even for long-time practitioners driving a car is a highly complex task. This becomes evident by the still high number of accidents.
    [Show full text]
  • Using Recurrent Neural Networks to Understand Human Reward Learning
    Using Recurrent Neural Networks to Understand Human Reward Learning Mingyu Song ([email protected]), Yael Niv ([email protected]) Princeton Neuroscience Institute, Princeton University Princeton, NJ 08540, United States Ming Bo Cai ([email protected]) International Research Center for Neurointelligence, (WPI-IRCN), UTIAS, The University of Tokyo Bunkyo City, Tokyo 113-0033, Japan Abstract Understandably, most cognitive modeling work focus on relative model comparison, which identifies the best-fit model Computational models are greatly useful in cognitive science in revealing the mechanisms of learning and decision making. out of a number of alternatives under consideration. Studies However, it is hard to know whether all meaningful variance do not, however, commonly evaluate the absolute goodness in behavior has been account for by the best-fit model selected of fit of models, or estimate how far they are from the best through model comparison. In this work, we propose to use recurrent neural networks (RNNs) to assess the limits of pre- possible model. In theory, there exists an upper bound for dictability afforded by a model of behavior, and reveal what (if model log likelihood (i.e., the negative entropy of the data anything) is missing in the cognitive models. We apply this ap- (Cover, 1999)). However, estimating this bound is practically proach in a complex reward-learning task with a large choice space and rich individual variability. The RNN models outper- impossible in sequential tasks, because the exact same expe- form the best known cognitive model through the entire learn- rience is never repeated (it may be possible, however, in sim- ing phase.
    [Show full text]
  • Cognitive Modeling for Intelligent Tutoring Systems
    Rule-Based Cognitive Modeling for Intelligent Tutoring Systems Vincent Aleven Human-Computer Interaction Institute, Carnegie Mellon University Pittsburgh, PA, USA [email protected] Abstract. Rule-based cognitive models serve many roles in intelligent tutoring systems (ITS) development. They help understand student thinking and problem solving, help guide many aspects of the design of a tutor, and can function as the “smarts” of a system. Cognitive Tutors using rule-based cognitive models have been proven to be successful in improving student learning in a range of learning domain. The chapter focuses on key practical aspects of model development for this type of tutors and describes two models in significant detail. First, a simple rule-based model built for fraction addition, created with the Cognitive Tutor Authoring Tools, illustrates the importance of a model’s flexibility and its cognitive fidelity. It also illustrates the model- tracing algorithm in greater detail than many previous publications. Second, a rule-based model used in the Geometry Cognitive Tutor illustrates how ease of engineering is a second important concern shaping a model used in a large-scale tutor. Although cognitive fidelity and ease of engineering are sometimes at odds, overall the model used in the Geometry Cognitive Tutor meets both concerns to a significant degree. On-going work in educational data mining may lead to novel techniques for improving the cognitive fidelity of models and thereby the effectiveness of tutors. Keywords: cognitive modeling, cognitive fidelity, intelligent tutoring systems, Cognitive Tutors, model-tracing tutors, authoring tools 1 Introduction Cognitive modeling has long been an integral part of ITS development.
    [Show full text]