This article describes an integrated theory of analogical access and mapping, instantiated in a

Advances in graphical technology have now made it possible for us to interact with information in innovative ways, most notably by exploring multimedia environments and by manipulating three-dimensional virtual worlds. Many benefits have been claimed for this new kind of interactivity, a general assumption being that learning and cognitive processing are facilitated. We point out, however, that little is known about the cognitive value of any graphical representations, be they good old-fashioned (e.g. diagrams) or more advanced (e.g. animations, multimedia, virtual reality). In our paper, we critique the disparate literature on graphical representations, focusing on four representative studies. Our analysis reveals a fragmented and poorly understood account of how graphical representations work, exposing a number of assumptions and fallacies. As an alternative we propose a new agenda for graphical representation research. This builds on the nascent theoretical approach within cognitive science that analyses the role played by external representations in relation to internal mental ones. We outline some of the central properties of this relationship that are necessary for the processing of graphical representations. Finally, we consider how this analysis can inform the selection and design of both traditional and advanced forms of graphical technology.

It is argued that working memory limitations are best defined in terms of the complexity of relations that can be processed in parallel. Relational complexity is related to processing loads in problem solving, and discriminates between higher animal species, as well as between children of different ages. Complexity is defined by the number of dimensions, or sources of variation, that are related. A unary relation has one argument and one source of variation, because its argument can be instantiated in only one way at a time. A binary relation has two arguments, and two sources of variation, because two argument instantiations are possible at once. Similarly, a ternary relation is three dimensional, a quaternary relation is four dimensional, and so on. Dimensionality is related to number of chunks, because both attributes on dimensions and chunks are independent units of information of arbitrary size. Empirical studies of working memory limitations indicate a soft limit which corresponds to processing one quaternary relation in parallel. More complex concepts are processed by segmentation or conceptual chunking. Segmentation entails breaking tasks into components which do not exceed processing capacity, and which are processed serially. Conceptual chunking entails "collapsing" representations to reduce their dimensionality and consequently their processing load, but at the cost of making some relational information inaccessible. Parallel distributed processing implementations of relational representations show that relations with more arguments entail a higher computational cost, which corresponds to empirical observations of higher processing loads in humans. Empirical evidence is presented that relational complexity discriminates between higher species...

Similarity-based and rule-based accounts of cognition are often portrayed as opposing accounts. In this paper we suggest that in learning and development, the process of comparison can act as a bridge between similarity-based and rule-based processing. We suggest that comparison involves a process of structural alignment and mapping between two representations. This kind

Abstract not found

This paper begins with a brief review of SME and MAC/FAC, our simulations of matching and retrieval. Next I lay out several arguments for exploring analogy in the large, including why it is now very feasible and what we can learn by such explorations. A new constraint on cognitive simulations, the Integration Constraint, is proposed: A cognitive simulation of some aspect of analogical processing should be usable as a component in larger-scale cognitive simulations. I believe that the implications of this new constraint for cognitive simulation of analogy are far-reaching. After that, two explorations of larger-scale phenomena are described. First, I describe a theoretical framework in which we model common sense reasoning as an interplay of analogical and first-principles reasoning. Second, I describe how SME and MAC/FAC have been used in a case-based coach that is accessible to engineering thermodynamics students worldwide via electronic mail. These examples show that exploring analogy in the large can provide new insights and new challenges to our simulations. Finally, the broader implications of this approach are discussed.

In order to allow robot agents to adapt their behaviour, a new approach -- learning by imitation -- has been proposed, in which a robot learns novel behaviours through interactions with the environment and other agents. In this paper we describe how our sighted agent, ESCHeR -- who is equipped with dual foveated lenses and can control head, neck and eye joints -- develops fine oculo-motor control through interaction with the environment. ESCHeR successfully learns to coordinate pan, tilt and vergence such that he can track bright moving objects and saccade rapidly to new objects of interest. In developmental terms, ESCHeR can be considered to have progressed from stage 1 to stage 2 of imitation learning. A novel representation of visual scenes is then introduced, and it is discussed how ESCHeR will use this to progress to stage 3. 1 Introduction For intelligent robots to behave within a dynamic world and interact with each other, an accurate perception of their environment is requir...

Inferences about spatial, temporal, and other relations are ubiquitous. This article presents a novel model-based theory of such reasoning. The theory depends on 5 principles. (a) The structure of mental models is iconic as far as possible. (b) The logical consequences of relations emerge from models constructed from the meanings of the relations and from knowledge. (c) Individuals tend to construct only a single, typical model. (d) They spontaneously develop their own strategies for relational reasoning. (e) Regardless of strategy, the difficulty of an inference depends on the process of integration of the information from separate premises, the number of entities that have to be integrated to form a model, and the depth of the relation. The article describes computer implementations of the theory and presents experimental results corroborating its main principles.

Standard feedforward and recurrent networks cannot support strong systematicity when constituents are presented as local input/output vectors (Phillips, 1998). To explain systematicity connectionists must either: (1) develop alternative models; or (2) justify the assumption of similar (non-local) constituent representations prior to the learning task. I show that the second commonly presumed option cannot account for systematicity, in general. This option, termed first-order connectionism, relies upon established spatial relationships between common-class constituents to account for systematic generalization: Inferences (functions) learned over, e.g., cats extend systematically to dogs by virtue of both being nouns with similar internal representations so that the function learned to make inferences employing one simultaneously has the capacity to make inferences employing the other. But, humans generalize beyond common-class constituents. Cross-category generalization (e.g., inferences that require treating mango as a colour, rather than a fruit) makes having had the necessary common context to learn similar constituent representations highly unlikely. At best, the constituent similarity proposal encodes for one binary relationship between any two constituents, at any one time. It cannot account for inferences, such as transverse patterning that require identifying and applying one of many possible binary constituent relationships that is contingent on a third constituent (i.e., ternary relationship). Connectionists are, therefore, left with the first option which amounts to developing models with the symbol-like capacity to explicitly represent constituent relations independent of constituent contents, such as in tensor-related models. However, rather just simply impl...

This paper examines human planning abilities, using as its inspiration planning techniques developed in artificial intelligence. AI research has shown that in certain problems partial-order planners, which manipulate partial plans while not committing to a particular ordering of those partial plans, are more efficient than total-order planners, which represent all partial plans as totally ordered. This research asks whether total-order planning and/or partial-order planning are accurate descriptions of human planning, and if different populations use different planning techniques. Using a simple planning task modeled after tasks designed in artificial intelligence we tested 7-- 8 year-old children, 11--13 year-old children, adult controls, and adults with damage to the prefrontal cortex. We found that adults and older children exhibited performance on planning tasks of varying complexity which matched that of artificial partial-order planners, and that this pattern of performance did not vary with multiple presentations of the planning task. In contrast, young children and adults with damage to the prefrontal cortex exhibited performance matching that of artificial total-order planners. This pattern of performance did vary, however, with multiple presentations of the planning task, with the young children and adults with cortical damage displaying aspects of total-order planning. In a further study we found that adolescents who had sustained damage to the prefrontal cortex as children displayed two