What are some key mechanisms of change in the information processing approach?

Publisher Summary

This chapter reviews a representative research generated by the information processing approach to human cognition and presents some implications of this research for the design of safe, comfortable, and efficient aviation systems. The information processing approach has not been without its critics. It has been accused of playing 20 questions with nature and losing. The essence of this criticism is that the science has become too segmented along the line of specific experimental tasks, that is, one group studies the infinite permutations on choice reaction time tasks, another group studies memory search tasks, and another focuses only on tracking tasks. Thus, the science has produced an enormous catalog of information about a few rather esoteric laboratory tasks; yet it has contributed very little to the understanding of how humans function outside the laboratory in real-world settings. It is likely that the future success of the information processing approach will rest on its ability to deal with this criticism. There is a need for human factors specialists to widen their perspective beyond the relatively simple tasks that currently dominate research to increase the complexity of experimental tasks and to incorporate more ecologically valid sources of information. The information processing paradigm has contributed both knowledge and tools relevant for understanding human performance in aviation systems. The study of human performance in aviation systems provides an excellent opportunity to better understand general issues related to human cognition in complex environments.

Publisher Summary

This chapter discusses the attentional demands of a simple voluntary positioning movement. It presents an understanding of motor control and performance that can be aided by an information-processing approach. The study of voluntary positioning movements is a venerable topic in experimental psychology with antecedents as early as the nineteenth century. A simple positioning movement requires an individual, sometimes called a subject (psychology jargon) or an operator (engineering jargon), to relocate a pointer such as a stylus or a finger by executing a spatial movement to reach a well-defined target position. The experimenter who induces such a voluntary movement is interested primarily in the speed and accuracy with which the pointer gets from here to there. Although terminology has changed, current views about simple positioning movements bear a strong resemblance to the position stated by Woodworth (1899). A ballistic or open-loop initial movement phase is followed by a closed-loop phase in which the processing of feedback information controls pointer position until the target is reached. The open-loop portion of movement is quite often said to be under the control of a motor program. Automatization releases attention or capacity so that it is available for performing a secondary task. Time-sharing paradigms are in some ways methodologically simpler than probe paradigms, they cannot sweep out time patterns of attention throughout a movement when analyzed.

Publisher Summary

This chapter discusses information processing model, cast in the form of a production system that could be used to explain different patterns of success and failure and the effects of training. At the heart of the production system were some things called quantification operators. These quantification operators were nice examples of what Walter Reitman meant when he once characterized the information processing approach as a way to invent what is needed to be known. They were essential to the logic of the model discussed in the chapter, and so their existence was postulated. The overall research strategy is to formulate models of performance of the developing organism at two different points in time, and then to formulate a model for the transition or developmental mechanisms. Adopting an extreme engineering approach to the information processing system, changes can be viewed in terms of four major classes of variables: (1) programs, (2) data structures, (3) capacities, and (4) rates. Changes in the first two result from software variation; changes in the last two result from hardware variation.

Publisher Summary

This chapter discusses many questions that arise when trying to understand the control processes involved in different forms of complex motor behavior. One convenient way to grasp a fairly wide perspective of task complexity is to view it from two of the principal approaches to understanding motor control. These two approaches, the “preprogramming model” and the “information processing” approach, may be thought of as being different dimensions, each capable of providing insight into the nature of task complexity. In essence, the notion of motor programming embodies the view that stored sets of motor commands, both innate and learned, are available within the central nervous system to be called upon at will and synthesized into a desired movement. The coordinative structures, the body's reflexes, are most familiar to clinicians who frequently see them manifest in the normal course of development of the neonate or in the motor expression of the brain-damaged and mentally retarded. Spinal reflexes such as the stretch reflex are readily identified but the higher level reflexes, for example, labyrinthine and righting reflexes, are less well known.

3.6 Human-Computer interaction

Human-Computer Interaction (HCI) is a subfield within computer science concerned with the study of the interaction between people (users) and computers and the design, evaluation and implementation of user interfaces for computer systems that are receptive to the user's needs and habits. It is a multidisciplinary field, which incorporates computer science, behavioral sciences, and design. A central objective of HCI is to make computer systems more user-friendly and more usable. Users interact with computer systems through a user interface, which consists of hard- and software that provides means of input, allowing users to manipulate the system, and output, allowing the system to provide information to the user. The design, implementation and evaluation of interfaces is therefore a central focus of HCI.

It is recognized in HCI that good interface design presupposes a good theory or model of human-computer interaction, and that such a theory should be based in large part on a theory of human cognition to model the cognitive processes of users interacting with computer systems [Peschl and Stary, 1998]. Such theories of human cognition are usually derived from cognitive psychology or the multi-disciplinary field of cognitive science. Whereas philosophers have rarely studied human-computer interaction specifically, they have contributed significantly to theorizing about cognition, including the relation between cognition and the external environment, and this is where philosophy relates to HCI.

Research in HCI has initially relied extensively on classical conceptions of cognition as developed in cognitive psychology and cognitive science. Classical conceptions, alternatively called cognitivism or the information-processing approach, hold that cognition is an internal mental process that can be analyzed largely independently of the body of the environment, and which involves the manipulation of discrete, internal states (representations or symbols) that are manipulated according to rules or algorithms [Haugeland, 1978]. These internal representations are intended to correspond to structures in the external world, which is conceived of as an objective reality fully independent of the mind. Cognitivism has been influenced by the rationalist tradition in philosophy, from Descartes to Jerry Fodor, which construes the mind as an entity separate from both the body and the world, and cognition as an abstract rational, process. Critics have assailed cognitivism for these assumptions, and have argued that cognitivism cannot explain cognition as it actually takes place in real-life settings. In its place, they have developed embodied and situated approaches to cognition that conceive of cognition as a process that cannot be understood without intimate reference to the human body and to the interactions of humans with their physical and social environment [Anderson, 2003]. Many approaches in HCI now embrace an embodied and/or situated perspective on cognition.

Embodied and situated approaches share many assumptions, and often no distinction is made between them. Embodied cognition approaches hold that cognition is a process that cannot be understood without reference to the perceptual and motor capacities of the body and the body's internal milieu, and that many cognitive processes arise out of real-time goal-directed interactions of our bodies with the environment. Situated cognition approaches hold that cognitive processes are co-determined by the local situations in which agents find themselves. Knowledge is constructed out of direct interaction with the environment rather than derived from prior rules and representations in the mind. Cognition and knowledge are therefore radically context-dependent and can only be understood by considering the environment in which cognition takes place and the agent's interactions with this environment.

Embodied and situated approaches have been strongly influenced by phenomenology, especially Heidegger, Merleau-Ponty and the contemporary work of Hubert Dreyfus (e.g., [Winograd and Flores, 1987; Dourish, 2001; Suchman, 1987]). Philosophers Andy Clark and David Chalmers have developed an influential embodied/situated theory of cognition, active externalism, according to which cognition is not a property of individual agents but of agent-environment pairings. They argue that external objects play a significant role in aiding cognitive processes, and that therefore cognitive processes extend to both mind and environment. This implies, they argue, that mind and environment together constitute a cognitive system, and the mind can be conceived of as extending beyond the skull [Clark and Chalmers, 1998; Clark, 1997]. Clark uses the terms “wideware” and “cognitive technology” to denote structures in the environment that are used to extend cognitive processes, and he argues that because we have always extended our minds using cognitive technologies, we have always been cyborgs [Clark, 2003]. Active externalism has been inspired by, and inspires, distributed cognition approaches to cognition [Hutchins, 1995], according to which cognitive processes may be distributed over agents and external environmental structures, as well as over the members of social groups. Distributed cognition approaches have been applied to HCI [Hollan, Hutchins and Kirsh, 2000], and have been especially influential in the area of Computer Supported Cooperative Work (CSCW).

Brey [2005] has invoked cognitive externalist and distributed cognition approaches to analyze how computer systems extend human cognition in humancomputer interaction. He claims that humans have always used dedicated artifacts to support cognition, artifacts like calendars and calculators, which HCI researcher Donald Norman [1993] has called cognitive artifacts. Computer systems are extremely versatile and powerful cognitive artifacts that can support almost any cognitive task. They are capable of engaging in a unique symbiotic relationship with humans to create hybrid cognitive systems in which a human and an artificial processor process information in tandem. However, Brey argues, not all uses of computer systems are cognitive. With the emergence of graphical user interfaces, multimedia and virtual environments, the computer is now often used to simulate environments to support communication, play, creative expression, and social interaction. Brey argues that while such activities may involve distributed cognition, they are not primarily cognitive themselves. Interface design has to take into account whether the primary aim of applications is cognitive or simulational, and different design criteria exist for both.

Architectural Design Modelling and Graphic Representations

Throughout the early stages of the design process, drawings are the main vehicle used by architects to externalize their current internal representations of the problem structure and of potential design solutions. Early design drawings often look like a private notation system which can only be communicated to others when supplemented by verbal commentary, what Schon (1983), (1985) has called the “language of designing.” As the design develops, and as tentative ideas are merged into a solution to which the designer is increasingly committed (through “shifts in stances,” in Schon's terminology), drawings become more explicit and capable of being understood by others on their own. The skilled designer will constantly monitor his or her own performance and select those design strategies that are most likely to lead to the desired solution (the “language about designing,” in Schon's terminology). Thus Schon emphasizes the designer's self-reflectivity while designing, what he calls “reflection-in-action,” and presents designing as a conversation with the materials of a “situation” (by venturing a design conjecture retrieved from memory). By remaining attentive to feedback, positive and negative, the designer is “reframing” problem situations, while also adapting situations to frames by exploiting new design opportunities hinted at by the current design. This procedure takes place through a sequence of “design experiments” which in contrast to scientific experiments, are all at once “exploratory, move-testing” and “hypothesis-testing.” During the process, representations such as drawings play the role of “virtual worlds” which enable the “accurate rehearsal” of intended actions. The designer's changing grasp of the problem and his or her production of a solution are thereby interlocked throughout the process.

This analysis of the design/drawing process can be paralleled to the classificatory framework proposed by Goumain (1973) for studying the architectural design process. In this framework, an information-processing approach is taken to analyze the design activity both at the level of the individual designer and at the level of the design organization. The sources of information used to carry out these activities can be traced to five levels of organizational complexity: (1) the individual designer's memory; (2) solution evaluations, drawings, and written lists produced by earlier design activities; (3) the design office library, technical literature, design manuals, handbook of office procedures, previous completed jobs, standard details, etc., as well as advice from professional colleagues within the office; (4) the design coalition which includes all “agents” who make a contribution to the design process (consultants, etc.); and (5) the external environment where the design will eventually be used. It includes the public at large, its democratically elected representative, and officers such as the town planner and building inspector.

When confronted with any new design task, the designer brings to bear the totality of his or her prior experience and knowledge to perceive and understand the nature of the task and to preselect tentative solutions. These observations are consistent with Lebahar's (1983) thesis that the architectural design process is one of uncertainty reduction through the use of graphic simulation and sketches. The graphic simulation model provides a temporary representation of the problem that allows hypotheses concerning the “object model” to be expressed, tested, and modified prior to reaching conclusions concerning these hypotheses.

In summary, it is through the constant exercise of modelling and simulating by drawing that the architectural designer develops the ability to think in three dimensions and to manipulate the symbolic codes and conventions of architectural representation by drawing. This spatial ability is also developed through the making, manipulating, and visualizing of physical three-dimensional models. Through increasing experience the designer can call upon a widening arsenal of design strategies. Having assimilated these in relation to analogous design problems, he or she can successfully accommodate them to reduce the uncertainty presented by new problems.

If CAAD systems are to become “intelligent,” they must incorporate an explicit knowledge of the semantics of architectural symbols and drawings. In doing so, they should not bypass the designer's learning and growth process. Clearly, it is the task of design research to provide such knowledge.

One approach to the study of the semantics of architectural representations is to focus on their memorization in relation to design experience. Mallen and Goumain (1973) together with Wood (1973) suggested that the experienced designer organizes the information content of an architectural drawing differently than the layman, much as chess grandmasters appear to structure information about positions on a chess board in different and larger chunks than ordinary players (Chase and Simon, 1973). In a laboratory experiment, subjects were asked to copy an original drawing by examining it for as long as they wanted (exposure time), then turning around on a swivel chair to transfer the information onto a drawing board (drawing time), and then repeating the process as many times as needed (Figures 8 and 9). Videotaped records and interviews permitted the identification of chunks of information memorized for every exposure/drawing cycle in the sequence. For the copying of an architectural plan, chunking was shown to be strikingly different in the experienced and inexperienced subjects. For the former, both exposure and drawing times were very lengthy for the initial cycles, while these times were evenly distributed across the drawing cycles for the latter. Experienced architects clearly brought their experience to bear. For example, one of them was convinced that he was copying the original drawing when in fact he was using his knowledge of plumbing to redesign the layout of a bathroom. A control experiment involved the memorization of a non- architectural original (a Mondrian painting) and showed almost identical patterns between experienced and inexperienced subjects. Akin (1986) reports a set of three similar experiments where subjects were asked to interpret, trace, or copy the plan of a church (pp.119-130). Findings broadly concur with the above. In particular, they appear to confirm the hierarchically organized and nested structure of memory chunks in architectural design.

Long-Term Memory

Consistent with recent emphasis in cognitive psychology, the papers at this symposium mention long-term memory (LTM) explicitly less often than they do STM. Nevertheless, they suggest a role for LTM in development that is different in one fundamental respect from the role accorded to it classically.

A vulgar view of LTM might picture it as a large bin or filing cabinet in which the child, in the course of his development, accumulates new facts and knowledge. I suppose no one would quarrel with the proposition that this is part of what happens. To acquire a logograph-reading skill, the child must learn the meanings of a suitable vocabulary of logographs. This is a matter of storing away in the filing cabinet properly indexed pairs of associates. However, we have seen that it is equally essential that the child acquire appropriate programs for processing the logographs in accordance with the “reading” instructions.

In almost all references to LTM in this symposium, the speakers talk about the storage of “programs” or “strategies” or “rules, rituals, and tricks of the trade”—that is, of processes rather than information. This, of course, is the touchstone of the information processing point of view in psychology: “Knowing” is largely “knowing how”—that is, skill. It is the viewpoint that led Bartlett, in Thinking (1958), to take motor skill as his metaphor for thinking ability.

The Genevans in this symposium, Inhelder and Cellérier, have some interesting points to make about the relation between a structuralist view, which describes the concepts that the child acquires as abstract structures, and an information-processing approach, which describes them as programs—or, in the terminology of Inhelder and Cellérier, as schemata (see Part III).

I am reminded by the structure-schema distinction of the analogous distinction that linguists make between language competence and performance. In the view of some, language competence formulated, say, as a transformational grammar, provides an abstract description of what the native speaker “knows,” but does not describe the form in which that knowledge is held in memory or used to process language. I hasten to add that I am not at all sure that Inhelder and Cellérier would accept this analogy with the structure-schema dichotomy.

In her descriptions of some experiments on length conservation, Inhelder offers an interesting hypothesis about what needs to be stored in LTM before the child can perform such tasks. The tasks involve constructing a line of some sort that matches in length a line presented by the experimenter. The lines, or “roads,” are constructed from matchsticks, the experimenter using sticks of a different length than those used by the child. Hence, as in all the classic conservation experiments, the situation confronts the child with conflicting cues: He can count matchsticks, or he can estimate lengths. The difficulty in the task, argues Inhelder, lies in reconciling the judgments of equivalence arrived at by these different routes, and choosing a criterion that is consistent with the requirements of the task. She offers a similar analysis of the standard matching test that involves comparing the sizes of sets and subsets.

Now this interpretation is clearly not intended by Inhelder to supercede the usual Geneva analysis, since she speaks explicitly of “resolution through reciprocal assimilation of two different subsystems that do not necessarily belong to the same developmental level.” Thus, underlying the learning phenomena are structures stored in LTM that are acquired at different stages of development. If that is so, then we must suppose that each of these structures is associated with (1) processes of attention and perceptual encoding for acquiring information relating to the structure (e.g., counting operations and length-estimating operations for visual stimuli); and (2) an internal representation for encoding and storing in LTM information characterizing the structure.

The work by Klahr and Wallace (Part IV) can be interpreted as an endeavor to make entirely explicit the information processing associated with these kinds of cognitive structures. These authors agree with the other symposiasts in filling LTM mainly with programs; but they detail not only the programs, but also the encoding of information in LTM—the nature of the internal representation. They postulate that such information is stored in the form of lists and description lists—the latter being better known to contemporary psychologists as “feature lists.” A description, or feature, is simply a two-termed relation between an object and one of its properties: e.g., the color (relation) of the apple (object) is red (property or value).

An interesting characteristic of this representation is that it makes the contents of LTM rather homogeneous in organization, independently of the sensory channel through which the information was acquired. Thus, a mental picture is made of the same stuff (list structures of features) as a mental symphony. Of course, it is only the form of organization they share in common; the specific relations encoded depend on sensory mode—the feature “red” must have been acquired through the eyes, and “interval of a fifth” through the ears.

Postulating this common organization for encoding stimulus information illuminates one of the central issues discussed (Part III) by Jacqueline Goodnow in her paper: the issue of intersensory correspondences of stimuli. Suppose that a sequence of sounds is encoded as a list:

Tap−pause−tap−tap−tap

Suppose, further, that the child has a list of pairs (associations) in LTM:

tap→circle; pause→space; tap2 after tap1→circle2 to the right of                                                                            circle1

Then, a relatively simple program will allow him to translate the aural stimulus into a visual one, which he can undertake to draw:

circle−space−circle−circle−circle

What the homogeneous coding explains is the possibility of anyone even finding meaningful the task of intermodal correspondence. It does not explain why the task may be difficult for children. In her paper, Goodnow shows us what assumptions are involved in supposing that the child “should know” what intermodal associations the adult has in mind (why not: tap, pause → large circle; tap → small circle? Why not, indeed?). She demonstrates that the child must acquire, and store in LTM, a whole host of conventions, many of them culture-specific, about the correspondences that are “appropriate.”

B CUING PROPERTIES OF SIMULATORS