Wayne D. Gray
Human Factors & Applied Cognition
George Mason University
George Mason University
m/s 3f5
Fairfax, VA 22030
Phone: (703) 993-1357
Fax : (703) 993-1359
Email:
gray@gmu.edu
http://www.hfac.gmu.edu/People/WGray/Wgray.html
errors, rule-based behavior, cognitive skill, computational cognitive models, ACT-R, skilled behavior
For Card, Moran, and Newell's (1983) purposes, it was enough to conclude that skilled behavior could be characterized "solely by the time to completion" (p. 147). The time required to make, detect, and correct errors could be ignored as merely adding a small, random increment to completion time. Although this may have been a fair characterization of the role of errors in the types of immediate-feedback, office automation (primarily word processing) tasks upon which their work was based, the role of errors in many other systems cannot be so easily dismissed.
Using a combination of empirical and analytic methodologies, the proposed research will attempt to explicate the origin, nature, detection, and correction of errors in rule-based tasks (tasks such as programming a VCR or using an ATM that do not require extensive problem-solving). This goal will be achieved if a computational cognitive model can be built that is capable of tracing the cognitive processes and producing the cognitive products found in the empirical data. Closely related goals entail extending this understanding to include a description of how features of artifact design (e.g., features that encourage reliance upon display-based reasoning) interact with cognitive processes to affect error origin, detection, and correction; as well as, how the nature of errors changes as a function of user familiarity with the software.
Related to these scientific goals are the engineering goals of developing tools for human-computer interaction (HCI) researchers and designers. For researchers, the goals include introducing an empirical methodology, the microgenetic approach (Siegler, 1991), to the HCI research community. Additional tools for researchers will be created by extending ACT-R (Anderson, 1993) (a major force in theoretical cognitive science) to the sort of cognition involved in HCI. For designers, building ACT-R models that can trace complex human-computer interactions is an important step towards replacing existing methods of analytic modeling (such as GOMS) with a computational technology that is more powerful and easier to use. More immediately, the current work should provide information as to how design decisions affect the probability of making a particular error and the type of error made.
Gray, W. D. (manuscript submitted for publication). The nature, detection, and correction of errors in a rule-based programming task with a display-based control structure.
Gray, W. D. (1995). VCR-as-paradigm: A study and taxonomy of errors in an interactive task. In K. Nordby, P. Helmersen, D. J. Gilmore, & S. A. Arnesen (Eds.), Human-Computer Interaction--Interact'95, (pp. 265-270). New York: Chapman & Hall.
Anderson, J. R. (1993). Rules of the mind. Hillsdale, NJ: Erlbaum.
Card, S. K., Moran, T. P., & Newell, A. (1983). The psychology of human-computer interaction. Hillsdale, NJ: Lawrence Erlbaum Associates.
Siegler, R. S. (1991). The microgenetic method: A direct means for studying cognitive development. American Psychologist, 46(6), 606-620.
General. Human Factors and Applied Cognition combines theoretical cognitive psychology with a concern for practical applications. We try to combine two maxims: "there is nothing so useful as a good theory" (Lewin, 1952) and "nothing drives basic science better than a good applied problem" (Newell & Card, 1985). We try to convince our students to see the world of cognitive factors as originating from a limited cognitive system attempting to use a given artifact to accomplish a given task. Understanding the interactions among the cognition--artifact--task triad is not simple but neglecting any leg of this triangle provides false solutions that do not generalize outside of the lab or the development organization.
Research in this area often combines tool building with theory testing. We attempt to test cognitive theory in the context of important and complex (real-world) tasks. Doing so often requires a concomitant development to better ways of mapping task performance onto human cognition and performance (e.g., see the Gray, John, & Atwood, 1993 paper listed above). It may require the construction of artifacts (or parts of artifacts, widgets) that enhance some strength or weakness of human cognition.
Specifics. So, what do I really do? Okay. My current passion is computational cognitive modeling; theories such as ACT-R and Soar that combine much cognitive theory into a cognitive architecture. It is hard to test complex ideas without a tool that allows you to control the complexity. I believe that computational cognitive modeling provides the type of tool needed to manage this complexity. With the NSF grant I am using this tool to look at the cognitive processes involved in the errors made while performing routine tasks. In other research I am trying to nail down the elusive construct of cognitive workload. Attempts to understand the confusing and contradictory literature on this topic has led us to postulate the existence of micro-strategies. Micro-strategies are differences in behavior at the 50-500 msec level that may have a profound effect upon the successful outcome of higher level tasks. For example, the experienced difficulty (cognitive workload) of flying a plane may be higher for one pilot versus another because of the adoption of inefficient micro-strategies. Micro-strategies are NOT innate, may be trainable, and are not usually under conscious control.
See the above two references by Card, et al., 1980 and Anderson, 1993.
Gray, W. D., John, B. E., & Atwood, M. E. (1993). Project Ernestine: Validating a GOMS analysis for predicting and explaining real-world performance. Human-Computer Interaction, 8(3), 237-309.
Gray, W. D., Young, R., & Kirschenbaum, S. S. (in press). Special issue on cognitive architectures and human-computer interaction. Human-Computer Interaction.
John, B. E., & Kieras, D. E. (1997). The GOMS family of user interface analysis techniques: Comparison and contrast. ACM Transactions on Computer-Human Interaction, 3(4), 320-351.
John, B. E., & Kieras, D. E. (1997). Using GOMS for user interface design and evaluation: Which technique? ACM Transactions on Computer-Human Interaction, 3(4), 287-319.
Newell, A. (1973). You can't play 20 questions wtih nature and win: Projective comments on the papers of this symposium. In W. G. Chase (Ed.), Visual information processing, (pp. 283-308). New York: Academic Press.
Newell, A., & Card, S. K. (1985). The prospects for psychological science in human-computer interaction. Human-Computer Interaction, 1(3), 209-242.
Rieman, J., Young, R. M., & Howes, A. (1996). A dual-space model of iteratively deepening exploratory learning. International Journal of Human-Computer Studies, 44, 743-775.
I have a general interest in the usability of novel interfaces. A novel interface may involve almost anything that can be done to input or output information (in any form) between a human and a computer; examples include, see-thru layers, force-joysticks, combining speech-recognition technology with point-of-gaze, and so on. Whereas I am NOT interested in building these widgets myself, I am interested in their use characteristics. A key interest would be in how a new input, output, or display widget can be used with existing widgets to build an artifact that allows the user to accomplish some task. IN THIS CONTEXT, I am interested in how the new widget pushes around (makes demands upon) human cognition; strategies that it might incline the user to adopt (e.g., if the same information is displayed in different ways, different [and differently effective] decision making strategies will be adopted), so on and so forth.