A Theory of Human-Computer Interaction

The Trouble With HCI

Human-Computer Interaction (HCI) as a discipline is drifting. It has no long-term objectives and no strategic direction. This would be forgivable if, like other disciplines, it had methods which were widely accepted as being likely to produce increases in understanding. But HCI has no such methods. It is not possible to say that HCI is a science, it is barely possible to describe it as an engineering discipline and yet we, as practitioners, are reluctant to relegate it to the status of a craft.

Part of the difficulty appears to be that it is hard to identify the “core” subject-matter of HCI. This is often acclaimed by practitioners as a benefit because it leads to the need for a multi-disciplinary approach which is considered to be a good thing. Yet without an identifiable subject-matter, it is difficult to identify what are the important issues in HCI, it is hard to say what an appropriate method for studying or practicing HCI would be and it is impossible to develop a theory of HCI.

This last is of peculiar importance. If we could produce a theory of HCI (whatever that might be), it would at the same time tell us what HCI is (as defined by the theory) and what its scope is. If we could say what the core subject-matter of HCI was, we would, at the same time, be proposing something like a theory of HCI. The problem of finding a theory of HCI is clearly what Rittel and Webber called a “wicked problem”. That is, finding a way to understand the problem is tantamount to finding a way to solve it. The way to tackle the problem then is to break the vicious circle by creating a theory of HCI. This would then seed a process of elaboration and refutation which would eventually end in there being a theory (or set of theories) of HCI which were acceptable to practitioners in the field—almost regardless of the quality of the original.

What Would a Theory of HCI Look Like?

The definition of a theory used here is a little idiosyncratic but it is one that should seem familiar to HCI people. I would like to propose that a theory is the set of rules that defines the behaviour of a system of conceptual objects and relationships. The set of conceptual objects and relationships itself, I will call a conceptualisation and we can imagine that there could be an arbitrary number of theories for any particular conceptualisation as well as there being an arbitrary number of conceptualisations for any field.

A Conceptualisation

A conceptualisation belongs to a field just as a theory does. It describes in detail the objects that are important to the field and the relationships between them. In the field of HCI we could imagine a conceptualisation involving objects such as “user”, “computer”, “input device”, “data” and so on, and relationships between them such as “display”, “use”, “enter”, etc.. Yet there is an arbitrary number of other conceptualisations for the same field. If one were to read the HCI literature and attempt to extract the conceptualisation used by the author in each separate paper, a large set of more-or-less complete, vague, overlapping and coherent conceptualisations would be the result.

To some extent, it will be possible to say that some conceptualisations are “better” than others and we can imagine dimensions along which to judge them such as completeness of coverage of the field, the degree of discrimination or granularity of the concepts and the extent to which practitioners in the field will agree with the conceptualisation. However, it is also true that some conceptualisations may be better than others for different purposes and that many may be equally good for the same purpose. Nevertheless, I feel that the lack of a clearly articulated common conceptualisation for HCI is a significant problem for the field. It hinders communication within and outside the community of practitioners and it effectively blocks the development of theory.

A Theory

A theory is the set of rules that governs the behaviour of the system described by the conceptualisation. In Newtonian mechanics, for instance, the three laws of motion are examples of such rules.

It is unlikely that a theory of HCI would ever have the simple elegance of a physical theory. The conceptualisation of HCI is, to begin with, far more complex than the conceptualisation of space and mass. Additionally, the individual concepts are themselves considerably more complex than those found in physics. Contrast typical HCI concepts such as “user”, “display” and “message” with concepts from Newtonian physics such as “mass”, “distance” and “time”. Nevertheless, it is still conceivable that such a theory could be built—even if the rules were far more elaborate and commonly involved the use of qualification and uncertainty.