Abstracting Situated Action:
Implications for Cognitive
Modeling and Interface Design
If perceptivity and curiosity are indeed critical for scientific progress, why don't we in the
behavioral sciences teach our students to keep their eyes peeled?
—Frans de Waal, The Ape and the Sushi Master (2001, p. 186)
Perhaps no event in the history of the social and behavioral sciences has prompted a greater interest in field-based observational studies of human cognition and behavior than the widespread introduction of information technology into the human ecology. Recall, for example, that the human–computer interaction (HCI) discipline was originally conceived by Card, Moran, & Newell (1983) as largely a laboratory science based on experimental psychology and cognitive modeling methods. It did not take long, however, before it became widely recognized that the psychological and design issues amenable to study with such techniques, although certainly important, where nevertheless relatively narrow in scope compared to the demands of understanding and supporting interaction with realistically complex computer applications, distributed networks, and the social, organizational and cultural factors surrounding their use (Carroll, 1991; Carroll & Campbell, 1989; Klein & Calderwood, 1991; Olson & Olson, 1991; Rasmussen, 1990; Suchman, 1987). In turn, these concerns have led to the development of diverse conceptual frameworks and methods focused on broadening the domain of inquiry in human–technology interaction, largely eschewing the confines of the laboratory and instead promoting the importance of field research (e.g., Beyer & Holtzblatt, 1998; Cole, 1996; Dourish, 2001; Hollnagel, 2003; Hutchins, 1995; Kirsh, 2001; Militello & Hutton, 1998; Nardi, 1996; Suchman, 2000; Vicente, 1999; Woods, 2003).
Currently, however, a somewhat tenuous relationship remains between the largely descriptive, field-oriented perspectives and methods advocated by many of the authors just mentioned (e.g., activity theory, distributed cognition, situated cognition, cognitive task analysis, and ethnomethodology, among others) and the normative task of providing concrete prescriptions for design (e.g., Hughes, Randall, & Shapiro, 1992; Vicente, 1999). To take one illustrative example, in a previous review of Suchman's (1987) insightful field study of why a photocopier's “help” system was far from helpful, I commented that the design implications of that research were far from clear (Kirlik, 1998). Although Suchman's study clearly demonstrated that user interaction with the copier's “intelligent” help system was problematic, little in the way of concrete, normative guidance for how one might have improved the copier's design resulted from that work.
In reality, my observations were far from original: They merely echoed the reactions of many of