CRITIQUE AND RESPONSE: Response to Maddox Critique

Abstract

We sincerely thank both the editor and Michael Maddox for the opportunity to participate in this collegial debate centered on our article, Longitudinal Study of the Effects of Ecological Interface Design on Skill Acquisition (this issue). As we hope to show, the human factors discipline can benefit greatly from such an open process. We and Maddox agree on many issues, which is comforting to us as researchers, given Maddox's practical experience in the nuclear industry. The remaining disagreements seem to arise from differences in terminology and in the perceived role of human factors in systems design. We address the former issue first by defining several concepts that are part of the foundation of our work. An abstraction hierarchy (Rasmussen, 1979) is a multilevel representation of the structure of a work domain defined by a means-ends relation between levels. In process control, five levels of representation have been found to be of use: the purposes for which the plant was designed (functional purpose), the mass and energy topology of the plant (abstract function), the generic functions that implement that topology (generalized function), the plant components that realize those functions (physical function), and the spatial location and appearance of those components (physical form). A detailed description of an abstraction hierarchy for the DURESS II system that we used as a testbed in our research can be found in Bisantz & Vicente (1994). Physical information describes the state of material components in a plant (e.g., pumps, heaters, valves). Functional information describes the state of the functions or purposes that those components are intended to satisfy, rather than of the components themselves. The higher levels of the abstraction hierarchy contain functional information, whereas the lower levels contain physical information. Note that the distinction between physical and functional information is not equivalent to the implementation distinction between data that can be sensed directly and that which must be analytically derived from sensed data. Physical information can usually be sensed directly (e.g., valve position), but functional information can either be sensed directly (e.g., flow rate) or derived (e.g., energy inventory). An ecological interface is one which is designed according to the principles of the ecological interface design (EID) framework, summarized in our article (for more details, see Vicente & Rasmussen, 1992). The content of an ecological interface is defined by an abstraction hierarchy analysis of the plant and thus includes both physical and functional information. The P + F interface in our study satisfies this criterion, whereas the P interface does not because it contains only physical information and the status of system purposes (the latter must also be included, otherwise the system cannot be controlled). Vicente & Rasmussen (1990) provided a detailed description of how the principles of EID were used prescriptively to design a slightly different version of the P + F interface for the original DURESS system. These definitions invite a reinterpretation of two of Maddox's statements. First, he states that at least one other study examined operator performance with an ecological interface in the process control domain. Although the paper he cited does indeed use the adjective ecological to describe its displays, those displays do not satisfy the aforementioned definition. It is not surprising that there is confusion on this point; the human factors field has yet to reach the state of maturity in which all researchers consistently use the same set of well-defined terms, regardless of theoretical orientation. Second, Maddox points out that the idea of presenting process information according to abstraction hierarchies is not a new one and that such interfaces were designed after the Three Mile Island accident. We agree with the first part of this claim; Rasmussen's (1979) abstraction hierarchy was first described in a seminal technical report published 17 years ago. …