Architectural building blocks as the locus of adaptive behavior

Architectural Building Blocks as the Locus of Adaptive Behavior Selection Alonso H. Vera (alonso.vera@nasa.gov) Carnegie Mellon University & NASA Ames Research Center, Moffett Field, CA USA Irene Tollinger (irene.tollinger@nasa.gov) Katherine Eng (keng@mail.arc.nasa.gov) NASA Ames Research Center, Moffett Field, CA USA Richard L. Lewis (rickl@umich.edu) Department of Psychology, University of Michigan, Ann Arbor, MI USA Andrew Howes (howesa@mac.com) School of Informatics, University of Manchester, Manchester UK Abstract behavior variants uncovered in an experimental study. In what is a significant contribution, they demonstrate that CPM-GOMS has the potential to be used as an explanatory tool. The demonstration that people make microstrategy- level adjustments is an important contribution, and in this paper we further explore that hypothesis with their data. Gray and Boehm-Davis found an average 150-millisecond difference in task time between two different subtasks of their button task. Both subtasks require clicking a target not initially visible. In one case, the target location is known while in the other, the target location is not. Based on this study and related work (O’Hara & Payne, 1999), they argue that users optimize by selecting the most efficient microstrategy. Microstrategies are expressed as groupings of cognitive, motor, and perceptual operators into behavioral units, such as move-click and click-move. Microstrategies are basically the same level of analysis as templates in the CPM-GOMS method (e.g., John & Kieras, 1996). CPM-GOMS templates have proven useful as a modeling method for making predictions about the time course of behavior (Gray, et al., 1993). However, this does not constitute evidence that microstrategies actually represent the strategic units selected during task execution. While microstrategies may be a useful construct for reasoning about behavior post hoc, there must be a theory of behavior composition in order to provide an explanatory account of what occurs in the head during task execution. We have developed models using Gray and Boehm- Davis’s microstrategies within Apex-CPM and the CORE architecture over the course of the last several years (John, et al., 2002; Vera, et al., 2004). The work described here is motivated by a desire to extend the CPM-GOMS approach presented in “Milliseconds Matter” as a consequence of working with their microstrategies at a very detailed level. Over the last decade, CPM-GOMS practitioners have struggled to develop a coherent theory of behavior composition from microstrategies as evidenced by the difficulty in teaching microstrategy interleaving in courses and tutorials (B. E. John, personal communication, February 9, 2005). Similarly, the present authors and others (e.g. Vera, et al., in press) have worked toward the generation of such a theory. Although there has been substantial success Historically CPM-GOMS has been used to predict total time for long stretches of behavior. In “Milliseconds Matter”, Gray and Boehm-Davis (2000) use CPM-GOMS to develop microstrategy variants with subtly different internal structure to explain differences observed in empirical data collected. They argue for microstrategies as the basic unit of adaptive behavior selection. While the microstrategies developed provide a good fit to the data, there is neither direct evidence for microstrategies as compared to other possible constructs nor an explicit statement of the theory underlying their construction. While the use of CPM-GOMS as an explanatory mechanism is a substantial advance, microstrategies have theoretical and practical limitations in terms of: microstrategies functioning as cognitive units, composition and structure, and dependency constraints. An alternative construct called an Architectural Process Cascade (APC) is proposed as the locus of adaptive behavior selection. An APC-based model of the Gray and Boehm-Davis button study task is presented to address the limitations of microstrategies. Introduction Starting with Card, Moran, and Newell (1983), GOMS methods were developed to generate a priori predictions of human performance on human-computer interaction tasks. The CPM-GOMS method (Cognitive Perceptual Motor GOMS) in particular is useful in modeling routine skilled performance or “extreme expertise” (John & Kieras, 1996) at a high resolution. The analysis decomposes interactive activity into basic behaviors (clicking buttons, typing words, etc.), which are then expressed as concurrent, interleaving streams of cognitive, motor, and perceptual operators. Historically, the GOMS methods have served as engineering models, producing approximations of performance at the level of detail chosen by the modeler, to influence system design and evaluation rather than as theoretical, explanatory models (Gray, John & Atwood, 1993; John, 1990). As Card, Moran, and Newell (1983) state, engineering models are “intended to help us remember facts and predict user-computer interaction rather than intended as a statement of what is really in the head” (p. 24). Gray and Boehm-Davis (2000) reverse this practice by using CPM-GOMS in “Milliseconds Matter” to provide an account of what is happening in the head with respect to