What We Talk About When We Talk About Context

Posted: February 12th, 2006 | No Comments »

Dourish, P. 2004. What we talk about when we talk about context. Personal Ubiquitous Comput. 8, 1 (Feb. 2004), 19-30

The goal of this paper is to explore the technical and social perspective in terms of context. Dourish shifts the attention of context from “a set of descriptive features of seetings” to “practices – forms of engagement with those settings”:

We assigned a central role to the meanings that people find in the world and the meanings of their actions there in terms of the consequences and interpretations of those actions for themselves and for others.


the focus of the design is not simply “how can people get their work done,” but “how can people create their own meanings and uses for the system in use”;

Context plays a central role in ubiquitous computing, because now that computation is moved “off the desktop”, then we need to keep track of where it has gone. However the use of context vary:

Ecoded or dynamic context
Encoded: Systems encode context along with information so that is can later be used as retrieval cue.
Dynamic: Use context dynamically to tailor the behavior of the system or its response patterns of use.

Interactive system design often rigidly fails to respond to the setting in which action unfolds; by incorporating context, system designers have hoped to make their system more responsive to the different social settings in which they might be used.

Situated action and improvisation
Based on Shuman’s notion of “situated actions”, computer systems should respond to the settings within they are used (citing Abowd in The Human Experience, Ubicomp‘s effort informed by a situation action also emphasize improvisational behavior and would not required, or anticipate, the user to follow a predefined script).

However the social and technical ideas often sit uneasily together:

Ubiquitous computing systems may be more responsive, and yet they seem to fail to address the sociological critique. Turning social observation into technical design seems to be problematic.

Positivist and phenomenological reasoning
Seek to reduce social phenomena to essences or simplified models that capture underlying patterns. It is a quantitative or mathematical perspective (engineering approach).

In particular, the idea that context consists of a set of features of the environment surrounding generic activities, and that these features can be encoded and made available to a software system alongside an encoding of the activity itself, is a common assumption in many systems. It is inherent in the notion that our systems will “capture,” “represent,” or “model” context – the normal and appropriate concerns of positivist design.

Phenomenological theories
Subjective (social facts are emergent properties of interactions) and qualitative in orientation.

… engineering approaches – including those that tend to dominate discourse about ubiquitous computing – inherit from a positivist tradition, while many approaches to social analysis relevant to HCI design – including the ethnomethodological position practiced by Suchman and cited by Weiser – are heir to a phenomenological legacy.

Dourish reconsiders context not as a representational problem but as an interactional problem and assumes an alternative view on context:

  • Contextuality is a relation property that holds between objects or activities
  • The scope of contextual features is defined dynamically
  • Context is an occasioned property, relevant to particular settings
  • Context arises from the activity (no separation of context and content)

He suggest that ubiquitous computing helps supporting the context that should not be predefined. A system should be able to display aspects of its own context – its activities and the resources around which the activity is organized and adapt to them. The “gulf of interpretation” (the difficulty of interpretation the system’s state as a response to the user’s command is one of HCI’s major problem.

On a side note, Dourish mentions the differrent terms used to describe the integration of computer technology with the everyday physical world: ubiquitous computing (Weiser, 1991), context-aware computing (Dey et al., 2001), pervasive computing (Ark and Selker, 1999), embodied interaction (Dourish, 2001), and more. A subject already covered in Disambiguating the Terminology around Ubiquitous Computing.

Relation to my thesis: I am also trying to bridge the technical and social perspectives and try to find ways to improve the missmatches. Context is hard to grasp and if not done properly leads to the gulf of interpretation. Current workarounds are about systems being adaptive and displaying their context (which lead to to other issues on the user’s mental load). Dourish focuses on “how can people create their own meaning and uses for the system in use” instead of the very engineering vision of “how can people get their work done”. My thesis should have this first approach (designing technlogies, but also chaning technologies that we design)

We can support the emergence and use of these structures, but we cannot separate them, analytically or technically, from the circumstances and occasions of their production.