interaction, artificial intelligence. People make use of hidden external information, first recalling that it exists and then finding it. This dissertation investigates the memory phenomena involved in recalling that external information exists. We present data in which a programmer navigates to hidden features in a real-world task environment. We then present a model that accounts for this navigation by encoding and using simple episodic memories for having seen a feature. The model inherits constraints from its underlying cognitive architecture, which specify that learning is passive and pervasive, and that it creates simple memories that depend on the feature itself being present as a cue. The nature of these memories requires the model to recall features to its mind’s eye as cues in order to retrieve them. This retrieval process requires domain knowledge: familiarity with features in order to imagine them, and an idea of when it would be useful to recall having seen them. Recalling that a hidden feature exists prompts the model to scroll to that feature. Thus the model’s access to external information is a function of passively-encoded episodic memories, and retrieval of these memories using knowledge. As a claim applied to people, this appears to overlap with a recently-

In this paper, we review Icarus, a cognitive architecture that utilizes hierarchical skills and concepts for reactive execution in physical environments. In addition, we present two extensions to the framework. The first involves the incorporation of means-ends analysis, which lets the system compose known skills to solve novel problems. The second involves the storage of new skills that are based on successful means-ends traces. We report experimental studies of these mechanisms on three distinct domains. Our results suggest that the two methods interact to acquire useful skill hierarchies that generalize well and that reduce the effort required to handle new tasks. We conclude with a discussion of related work on learning and prospects for additional research, including extending the framework to cover developmental phenomena. Key words: incremental learning, cognitive architecture, reactive control, problem solving, hierarchical skills

This paper describes findings from a study conducted to analyze the behavior, knowledge, and thinking that support the highest levels of human land mine detection skill. A recent assessment of land mine detection capability concluded that “human operators perform better with any detector system than the corresponding fully automated system. ” 1 This assessment, plus evidence linking individual differences in detection ability to experience, suggests that methods, data, and theory developed in studying human expertise can be applied to the problems of land mine detection and discrimination. Studies of experts across a variety of skill domains have demonstrated that analyses of experts ’ skills can yield findings useful for designing efficient and effective training programs and supporting technology development. This initial field study was performed to (a) identify the upper levels of human mine detection capability using currently-fielded hand-held equipment and (b) model the knowledge, thinking, and techniques employed by proficient human operators. Two experienced operators showed sufficiently impressive detection performance to qualify as experts. Data also show that a skilled PSS-12 operator can detect low-metal mines with considerable accuracy. A first-approximation information-processing model of expert operator skill is presented that is based on observation of the operators ’ activities as they searched for mine targets.

adoption of the World Wide Web over the past decade by both content producers and information consumers has resulted in greater volumes of content on an ever-widening array of subjects to be available to larger numbers of users. Clearly we are past the stage at which the primary challenge is being able to put the information where the public can have access to it; now the challenges include making it possible for the user to find the information she wants, and contributing to the user’s ability to make sense of the information she finds. The search engine companies (Google being the most notable and successful) are working to improve the process of finding relevant information; the area that is the concern of this review has to do with what the user does with the information after it has been found. In particular, what are the processes involved with integrating information from a number of sources into some coherent piece of knowledge? What happens if the information from two different sources is contradictory? How does the user know when it is appropriate to make comparisons between information gathered from more than one source? And, at the crux of the matter, are there things we as information scientists can

ends on the latter instead of the former. Moreover, models may be underconstrained and are applied to modeling data post hoc without a sufficient consideration of whether there are patterns of data that the model could not produce (see Roberts & Pashler, 2000). Finally, models are often put forward as theories in an of themselves, without sufficient analysis and explanation of why the model accounts for the data (see McCloskey, 1991). In general, computational modeling would seem to be most productive when it is carried out in the service of clarifying theoretical claims and when it it tightly integrated with corresponding empirical studies. A number of formalisms have been applied to modeling cognitive processes, including production systems, discrimination nets, exemplar-based models, and connectionist models. The current chapter presents a brief summary of each of these approaches along with some illustrative examples. Particular emphasis is place on conne

What is the role of a cognitive architecture in shaping a model built within it? Compared with a model written in a programming language, the cognitive architecture offers theoretical constraints. These constraints can be "soft", in that some ways of constructing a model are facilitated and others made more difficult, or they can be "hard", in that certain aspects of a model are enforced and others are ruled out. We illustrate various of these possibilities. In the case of Soar, its learning mechanism is sufficiently constraining that it imposes hard constraints on models constructed within it. We describe how one of these hard constraints deriving from Soar's learning mechanism ensures that models constructed within Soar must learn a display-based skill and, other things being equal, must find display-based devices easier to learn than keyboard-based devices. We discuss the relation between architecture and model in terms of the degree to which a model is "compliant" with the constraints set by the architecture. Although doubts are sometimes expressed as to whether cognitive architectures have any empirical consequences for user modelling, our analysis shows that they do. Architectures play their part by imposing theoretical constraints on the models constructed within them, and the extent to which the influence of the architecture shows through in the model's behaviour depends on the compliancy of the model. 1.

Abstract not found

Latent Problem Solving Analysis (LPSA) is a theory of knowledge representation in complex problem solving that argues that problem spaces can be represented as multidimensional spaces and expertise is the construction of those spaces from immense amounts of experience. The model was applied using a dataset from a longitudinal experiment on control of thermodynamic systems. When the system is trained with expert-level amounts of experience (3 years), it can predict the end of a trial using the first three quarters with an accuracy of.9. If the system is prepared to mimic a novice (6 months) the prediction accuracy falls to.2. If the system is trained with 3 years of practice in an environment with no constraints, performance is similar to the novice baseline.

Landmines pose a major military threat as well as a serious humanitarian problem. The difficulty of detecting buried landmines, especially modern landmines with minimal metallic content, contributes heavily to the problem. The two projects described here took a cognitive engineering approach to improve detection capability via development and implementation of scientifically principled operator training. Each analyzed and modeled the landmine detection skills of expert users of two different handheld detection systems, the PSS-12, the U.S. Army standard equipment, and the PSS-14 1, an advanced technology system under development at the time of this work. Model-based training programs were developed and tested. Both proved highly effective for developing detection skills, producing the greatest increases in detection rates against the most challenging targets. The US Army has adopted and now uses both programs. Results demonstrate the practical utility of information-processing models of expertise for designing instruction and developing important human skills. 1.

Artificial intelligence methods may be used to model human intelligence or to build intelligent (expert) computer systems. Al has already reached the stage of human simulation where it can model such "ineffable " phenomena as intuition, insight and inspiration. This paper reviews the empirical evidence for these capabilities. 1