What is "context", anyway

A major line of research in ubiquitous computing is "context-awareness", which means computational systems can sense and respond to various aspects of the settings where we live and work. Attractive as it might be to have sensors and computational systems to take care of somethings for us, more fears and doubts are aroused considering how much privacy we have to sacrifice for it. Do we really want computers to be able to figure out our needs and even emotions for us? If that is not the direction we want to drive - to sacrifice our privacy for automation, then where is the direction?

Advancement of technology and desirable pictures of future life should form a symbiotic relationship. On one hand, a digital future with computation embedded in the fabric of the world is driven by the development of low power, low cost, ever smaller and ever smarter devices; on the other hand, the advocate of these pictures should also foster the development. It seems most literatures around ubiquitous computing simply emphasize the former, while do not give the sufficient attention to the later as it deserves. Part of the reason can be traced back to when the notion of "ubiquitous computing" was coined by Mark Weiser. He proposed "transparent computing", where users only need to know what to do, with little concerns about how to do. Thus, the computing should be smart enough to figure out our situations and do things for us with our least attention. While "transparent" is a nice concept to describe how easy it is to use computers to commit our tasks, its implication that computings can fetch our personal info without our attention is a different story. How should we resolve this ambivalence around ubiquitous computing? How should we make the technological development and the vision of its usage benefit each other?

(The following arguments are mainly based on Dourish's paper "What we talk about when we talk about context")
One approach to draw a desirable picture is to gain insights from social science point of view. Actually, according to Dourish's paper "What we talk about when we talk about context", the incorperation of "context" or "situation" in computational design is a repsonse to social scientists' critique of traditional interactive systems. Social scientists have argued that traditional interactive system design often fails to take into consideration the specific setting where action unfolds, and Suchman's notion of "situated actions" becomes a common souce for reponsive and sensitive computational systems. However, it seems to make responsive and sensitive computation do not address the sociological critique successfully, and "the social and technical ideas often sit uneasily together". Obviously there are problems to translate ideas between these two intellectual frames: computer science and social science.

The difficulties of turning social observation into technical design come from the incompatibility of the respective approaches adopted by these two disciplines. Computing system designers take the positivist's tradition, the rational, empirical and scientific tradition, seeking to reduce the complex observable phenomena to underlying idealized mathematical descriptions, reduce social phenomina to essences, or simplified models that capture underlying patterns, and seeking for objective, independent descriptions of social phenomena, abstracting from the detail of particular occastions or settings. On the other hand, social scientists' approach are heir to a phenomenological legacy; in this view, social facts are emergent properties of interactions, not pre-given or absolute but negotiated, contested, and subject to continual process of interpretation and reinterpretation. They turn the attention away from the idea of a stable external world as unproblematically recognized by all, and towards the idea that world is essentially a consensus of interpretations.

With these two distinct approaches, they have different interpretations of "context". System designers take a representational approach, concerned with how context can be encoded and represented. The concept of context in this perspective have 4 assumptions: context is a form of information - it is something that can be known and hence encoded and represented; context is delineable - so we can define what is context in advance; context is stable - the determination of the relevance of any potential contextual element can e made once for all; context and activity are seperable. However, this definition of context is not the kind of context in social scientists' minds. They take an alternative point of view: context is not information, but a relational property that holds between objects or activities - the question is not whether something is context or not, but whether it may or may not be contextually relevant to some particular activities; the scope of contextual features is defined dynamically; context is an occasional property, relevant to a particular setting, particular instance of action, and particular parties to that action; and context arise from activities. So we should not consider context as a representational problem, in which the question is about "what is context and how can it be encoded". Rather, we should take it as a interaction problem, and ask "how and why in the course of their interactions, do people achieve and maintain mutual understanding of the context for their actions". So just like ordinary-ness, context is also an "interactional" problem.

Since the main concern of the use of context is to discriminate or elaborate the meaing of user's activity, the link between aciton and meaing as a primary concern, and we find the link in "practice". According to Etienne Wenger, practice is not just what we do, but what we experience in the doing. It is a process by which we can experience the world and our engagement with is as meaningful. The unification of action and meaning is also central to the question of context, since context is essentially about the ways in which actions can be rendered as meaningful - how a particular action, for example, becomes meaningful for others by dint of where it was performed, when, or with howm. So what constitutes context, how people may orient to features of the world as contextual or central, how relevance is managed, etc. - are questions of practice.

What is crucial to the interactional view is to see practice as a dynamic process - it evolves and adapts. By turning attention from context as a set of descriptive features of settings to practice as forms of engagement with those settings. Thus, the concern of ubiquitous computing is not a "transparent" computing interface where computing systems take care of everything automatically, rather a open-ended new form of interaction where users engage, adopt and adapt technologies and incorporate them into their own work, and new meanings are created and communicated.