Olav W. Bertelsen
Department of Computer Science
University of
Aarhus
http://www.daimi.au.dk/~olavb
olavb@daimi.au.dk
I have been concerned for some time over the role of theory and theory production in human computer interaction, especially in conjunction with the concept of design artefacts (Bertelsen 1998).
Card et al. (1983) aimed at creating an engineering psychology as an approximate version of the general psychological theories; this psychology would be applicable rather then true. Their book was based on the idea that basic research, in a natural science way, produces true and universal theories, and that these theories later are turned into instruments for practitioners. Card et al. saw this as the only alternative to unsystematic design by rules of thumb and ad hoc procedures. However, this view on the role of scientific theory does not fit with the actual history of technology and it doesn't fit with basic features of human-computer interaction. A general lack of theories that are pragmatically rooted in praxis has been symptomatic for the field. This lack has not been solved by the ethnomethodological case studies, which tend to rejects theory in general.
The question is what kind of knowledge about human-computer interaction we can option. Can we treat it as a natural science? What will we use the results for? Loosely based on the mainstream of newer human-computer interaction research it is fair to say that the field has the following basic features. Human-computer interaction is societally constituted, i.e. socio-cultural insights are needed to understand what goes on in the interface. Human-computer interaction does not go on in isolation. The use quality of the interface is constituted in use. Computer artefacts, mediate purposeful action. The interface is in constant development, changing appearance as the user and use context develops. Interfaces are designed, interaction planned by others. These features imply that HCI cannot be studied isolated in a lab. Results often cannot be reproduced. The "validity" of "controllable" science most often comes at the expense of relevance. In short this means that we need a closer coupling or even merging between the theory/research loop and the practice loop.
In the same way as computer artefacts and interfaces are mediators of work, theories and other results of human-computer interaction research can be understood as mediators of design and research. In exploring how scientific theories work as design artefacts, I look at Fitts' law (Fitts 1954).
Theories play different roles in design; from worldviews, guiding the designer and helping him assess the situations and keep the goals in mind, to tools mediating the achievement of specific results.
In (Bertelsen, 1994) the notion of theories as design artefacts is introduced through the example of Fitts' law. Based on the literature, three distinct "design artefact roles" played by Fitts' law is identified. These are basic worldview, tool for calculation, and metaphor (source of inspiration). The borders between the three roles are not clear-cut. By thinking about interface problems in Fitts' law terms (the metaphor role), other views are excluded and Fitts' law draws the main perspective on the interface towards time performance concerns. As a tool for calculation Fitts' law is a primary artefact, just as a microscope. Of course, the things we are able to see by using Fitts' law are irrelevant for design unless the calculations can be related to the use situation. As a metaphor, Fitts' law is a tertiary design artefact. As part of the basic worldview, Fitts' law promotes the idea that human-computer interaction can be decomposed and subjected to universally valid laboratory experiments. In the later role, Fitts' law and the entire cognitive science framework is a secondary artefact.
The three roles of Fitts' law as a design artefact cannot be generalised. However, they span a spectrum of ways in which theories mediate design. Between being tool, which is goal directed and "neutral", and being worldview which is value based and motive oriented.
A theoretical complex like Activity theory takes a tool role when e.g. analysing empirical data of interaction with a specific computer artefact in terms of focus shifts (Bødker 1991). If, at the other extreme, designers adopt the activity theory framework as their way of understanding technology, i.e. as culturally constituted mediation of human activity etc., then activity theory takes the worldview role. In between these roles the well-known double triangle figure can serve as a vehicle for communication. In the human-computer interaction community in general "activity" has become a placeholder for the whole spectrum of insights about design as something more than mere time and motion optimising (see Blumenthal 1995, for an pronounced example of this use of activity theory).
Theories mediate construction, communication and conception by playing different roles in the continuum from tool to worldview, being primary, secondary and tertiary artefacts. Theories explicate a doubleness of design artefacts as mediating construction and representation; as in the case with Fitts' law both mediating technical optimisation of user interface layout, and representing specific understandings of the interface design problem. Theories, like other design artefacts, are boundary objects (Star 1989), by tying the production of "universal facts" in research together with design.
The notion of theories as design artefacts contributes to a pragmatic philosophy of design science by providing a criterion for the assessment of theory. The validity of a theory is appreciated based on its mediation of design activity, not just based on random preferences of individuals, but based in concrete societal praxis at a specific point on the trajectory of cultural development.
Taking seriously that the validity of theories can only be assessed through the praxis they induce, influences how research should be done. Cooperation with industrial practice is necessary, not only as a test bed for new ideas and a source for empirical data, but more importantly as tightly coupled partners in the course of research. Thus, in our project on computer support for wastewater treatment work (Bertelsen & Nielsen 1999), we have co-operated closely with a strategic task force at a Danish producer of mechatronics. This has given us easy access to knowledge about the specific domain, expanded our debate forum and given us a sound sense of reality.
However, the danger is that "the real world" is so fascinating in its enormous complexity that researchers easily become engulfed in making concrete solutions. Then, it becomes impossible to generalise and to produce theory. Thus, it is important to take time for reflection, to write purely theoretical papers and to make cruel reductions of the domains we well know are "more complex than that".
Bertelsen, O. W. (1998). Elements to a theory of design artefacts: a contribution to critical systems development research, Ph.D.-Thesis, Aarhus University. DAIMI PB-531.
Bertelsen, O.W. & C. Nielsen (1999). Dynamics in Wastewater Treatment: A Framework for Understanding Formal Constructs in Complex Technical Settings. In Proceedings of ECSCW99, Klüwer.
Blumenthal, B. (1995). Industrial Design and Activity Theory: A New Direction for Designing Computer-Based Artifacts. In Blumenthal, Gornostaev, & Unger (eds.). Human-Computer Interaction. 5th. International Conference, EWHCI `95 Moscow, Russia, July 1995. Selected Papers. Berlin: Springer Verlag (LNCS 1015). pp. 1-16.
Bødker, S. (1991). Through the Interface: a human activity approach to user interface design. Hillsdale, N.J.
Card, S. K.; Moran, T. P. & Newell, A. (1983). The Psychology of Human-Computer Interaction. Hillsdale NJ.
Fitts, P. M. (1954). The information capacity of the human motor system in controlling the amplitude of movement. In Journal of Experimental Psychology vol. 47, no. 6. pp. 381-391.
Star, S. L. (1989). The Structure of Ill-Structured Solutions: Boundary Objects and Heterogeneous Distributed Problem Solving. In Gasser, Les & Michael N. Huhns Distributed Artificial intelligence, volume II. London: Pitman Publishers, pp. 37-54.