Proceedings of the Annual Meeting of the Cognitive Science Society

In this paper we present a framework for describing cooperative work in informal domains such as an office. We argue that standard models of such work are inadequate for describing the adaptibility and variability that is observed in offices, and are fundamentally misleading as metaphors for understanding the skills and knowledge needed by computers or people to do the work. The basic claim in our alternative framework is that an agent's work is defined in terms of making and fulfilling commiunents to other agents. The tasks described in those commitments are merely agreed upon means for fulfilling the commitments, and the agents involved in the agreement decide in any given situation how and whether a given commitment has been fullfilled. We also claim that in informal domains, descriptions of tasks, functions, and procedures are necessarily incomplete and imprecise. They result from negotiations among the agents and serve as agreed upon specifications of what is to be done. The descriptions evolve during their use in continuing negotiations as situations and questions arise in which their meaning is unclear. These claims imply that for a given situation an agent using such descriptions must be capable of interpreting imprecise descriptions, dctcmiining effective methods for performing tasks, and negotiating with other agents to determine task requirements.

Everyday problem solving is different in significant ways from the kinds of problem solving that take place in laboratory mlcroworld settings. Attempts to simplify have excluded important factors that can help us understand aspects of the problem solving that are problems from the point of view of the laboratory. This paper describes several research projects that have examined problem solving in nonlaboratory settings, and some of the Implications of these studies for cognitive science. The current notions of distributed cognitive processing can be extended in a powerful way to the socially distributed problem solving characteristic of everyday settings. This notion of socially distributed problem solving can then reflect back on individual problem solving, which is acquired and often carried out in social settings.

In this paper we will give our reasons for believing that certain current attempts to explain perceptual phenomena on a lower level in terms of known sensorymechanisms are untenable. We will do this by focussing on two topics, lightness perception and the perception of apparent motion. We will summarize some older data (not all of which are sufficiently known) and will describe some recent work of our own. Finally, on a more positive note, we will try to indicate the direction that a theory must take if it is to deal effectively with these phenomena.

In this paper, we discuss the role of experiences and examples in learning systems. We discuss these issues in the context of three systems in particular: Rissland and Soloway's Constrained Example Generation (CEG) System, Selfridge's COUNT, and Soloway's BASEBALL.

Two studies run simultaneously investigated the influence of positive affect on creative problem solving as indicated by Duncker's (1945) candle task. Results show that positive affect, as induced by exposure to a funny movie, facilitated a subject's ability to solve the problem in comparison with those in control groups who either saw a control film or who did not view a film at all. In addition, in accord with previous findings (AJamson, 1952; Higgins & Chaires, 1980) , subjects in another comparison group who were exposed to a facilitative display of the items were also more likely than a control group to solve the candle task. Results are discussed in terms of the influence of a positive affective state on accessibility of material and on cognitive organization.

We present a syntactic parser. Paragram. which tries to accomrnodate three goals. First it will parse, in a natural way, ungrammatical sentences. Secondly, it aspires to 'capture the relevant generalizations'", as in transfornalional grammar, and thus its rules are in virtual one-to-one correspondence with typical transformational rules. Finally, it promises to be reasonably efficient, especially given certain limited parallel processing capabilities.

Predictions were made about the types and the number of inferences to be found in the verbal protocols of subjects reading difficult to understand texts. The predictions were made on the basis of two alternative models of inference generation, the backward bottom-up inference and the forward bottom-up inference. It was also hypothesized that plan-goal and script inferences would be stored longer in STM than coreference and role identification inferences. This implies that plan-goal and script inferences are more likely to be reported than coreference and role Identification inferences. The protocol analyses support the backward bottom-up inference model and support the assumed length of storage in STM of the different types of inferences.

This essay conjectures that an author's planning process will be facilitated by a tool that represents his plan at various levels of abstraction as a network of subgoals, with the subgoals not necessarily restricted to a linear order. Machine reasoning on such structures has been explored in artificial intelligence research; our proposal is to make these structures available to the writer as a calculus for representing his essays and to use the computer

Results in cognitive psychology have shown that readers can be steered away from an otherwise plausible interpretation of a story by extra-textual factors such as the source of the text, the stated reading purpose, interruptions and repetition of questions about the text. For instance, successive repetitions of the same question about a given text will often elicit a series of alternative interpretations of the text. This effect cannot be accounted for by established principles of text processing behavior, such as people's preference for cohesive and parsimonious representations of text. This paper presents a computer program called MACARTHUR, which models this behavior by varying the depth and direction of its inference pursuit in response to re-questioning, resulting in a series of markedly different interpretations of the same text. In light of the results, some new experiments are suggested in hopes of arriving at a new principle, beyond cohesion and parsimony, to account for the observed text processing behavior.

A computational model is proposed for the early stages of development of the object concept in infants. Stages in development are represented as a sequence of grammars or rewrite rules parsing a set of perceptual phenomena. The infants' changes between developmental stage; can be described by differences between the grarrmar rules modelling each stage. The program replicates five Bower et al studies on the development of the object concept and reaffirms the primacy of rest and motion parameters as explanatory invariants in early object concept development.

Design is concerned with how things ("artifacts") ought to be in order to attain goals and function. To gain a deeper understanding of the nature of design, we have studied design processes in several different domains; software engineering, creative writing, building models with technical kits and constructing learning environments.

Lehnert (1980) proposed a model thematic knowledge structures called 'plot units', which are structurally deflned sequences of mental states, positive events, and negative events In a clustering experiment, subjects were asked to sort 36 stories into groups. These groups were labeled by the subjects, and that data used to identify the nature of each mental category. Plot units generally provided a good fit to the clustering patterns in the data, with higher level clusters corresponding to discriminations on the nature of the outcome and judgements about the 'fairness' of the protagonist.

We review the major phenomena of skilled typing and propose a model of the control of the hands and fingers during typing. The model is based upon an Activation-Trigger-Schema system in which a hierarchical structure of schemata directs the selection of the letters to be typed and, then, controls the hand and finger movements by a cooperative, relaxation algorithm The interactions of the patterns of activation and inhibition among the schemata determine the temporal ordering for launching the keystrokes To account for the phenomena of doubling errors, the model has only "type" schemata -- no "token" schemata - with only a weak binding between the special schema that signals a doubling and its argument. The model exists as a working computer simulation and produces an output display of the hands and fingers moving over the Keyboard and it reproduces some of the major phenomena of typing, including the interkeypress latency times, the pattern of transposition errors found in skilled typists, and doubling errors. Although the model is clearly inadequate or wrong in some of its features and assumptions, it serves as a useful first approximation for the understanding of skilled typing.

Any process model of problem solving trying to capture goal directed progress needs an evaluation function of the states. Maximation of its gradient at each step serves as a decision rule in chosing among different legally possible moves. The present analysis qives sufficient conditions for the existence of an evaluation of states which is applicable to customary laboratory problems in the study of problem solving. The evaluation, moreover, is shown to establish a foundation for the measurement of "Insight", the anticipatory aspect of human problem solving capacity.

The understanding of simple, narrative episodes in which a protagonist successfully realizes a goal through a sequence of actions is studied. In two experiments, subjects rated the acceptability of sentences of the form "The protagonist does ACT J^ in order that the protagonist could ACT 2", where ACT 2 and ACT }_ were actions from the episode. Ratings were predicted by (I.e., inversely related to) distance within a narrative representation which organizes actions into sequences (action chains) reflecting aspects of the problem-solving plan employed by the protagonist. Subjects separated action chains that had been interleaved in a text. Mishap, irrelevant, and restorative actions were not incorporated directly into an attempt structure. Corrective actions, undoing the ill effects of mishaps, were incorporated. Further research is suggested.

Two experiments investigated the effects of the number of interesting events in a story on memory for that story. The results differed depending on whether or not the story followed a stereotypical script. For script-based stories, adding interesting events spaced throughout the stories decreased the stories' memorability. However, for non-script-based stories, adding a few interesting events aided memory for the stories, but adding too many decreased memory. These results were interpreted in terms of a limited resource model of story understanding limited resource model of story understanding that focusses processing on interesting events.

We examined the reasons people give for the actions in routine events and classified them as action enablement, main goal satisfaction, or external. We argue that the external reasons indicate that some actions in events are stored at more general lev- els in memory than the specific event schemas

Interpreting metaphors is an integral and inescapable process in human understanding of natural language. The current investigation analyzes analogical mappings underlying metaphors and their implications for inference and memory organization. Regularities have been observed indicating that certain types of conceptual relations are much more apt to remain invariant in analogical mappings than other relations, resulting in an induced invariance hierarchy. The central thesis is that human inference processes are governed by the same analogical mappings manifest as metaphors in language.

Subjects were presented real-world objects at varying quired to name then. Naming with the angular departure o from the object's canonical or suits suggest that, at least variant features that uniquely are not present, subjects internal representation of the nonical orientation in order ognition process.

Three demonstrations are presented and used to support a number of apparently unrelated claims about the internal rejresentations that people have when they perceive or imagine a spatial structure. The first demonstration illustrates properties of the spatial working memory that enables us to integrate successive glimpses of parts of an object into a coherent whole. The second demonstration shows that our ability to generate a mental image is severely limited by the form of our knowledge of the shape of an object. The third shows that the shape representation which we create when we attend to a whole object does not involve creating the kinds of shape representations for the parts of the object that we would form if we attended to them and saw them as wholes in their own right. The real motivation for this medley of demonstrations and for the interpretations offered is that these phenomena can all be seen as manifestations of a particular kind of parallel mechanism which is described briefly in the last section.