Context-Aware Computing Applications

One-line summary: A context-aware system is one that can determine and react to the current physical and computing context of mobile users and devices, by altering the information presented to users or commands issued by and on behalf of those users.

Overview/Main Points

• Context-aware computing focuses on mobile-computing in which mobile people and not just mobile computers are considered. A context-aware system adapts according to location of use, neighboring entities, accessible devices, and changes to all of these.
• Four categories of context-aware applications are presented as two orthogonal axis with two points each: fetching information vs. giving commands, and manual vs. automatic actions.
  • Proximate selection: this is the manual fetching of context-aware information about input/output devices, non-physical objects and services, or locations. UI issues of how to emphasize information based on degree of proximity exist.
  • Automatic contextual reconfiguration: this is the automatic selection of information or altering components based on context. For example, a white-board app may bring up new or existing pages when a person enters a new room, or an OS may decide to spin down a disk when AC power is disconnected.
  • Contextual information and commands: queries on information or commands themselves may be altered by the context of the user. For example, a migrate button could bring a user's X display to a whiteboard in that user's current room. Or, a whiteboard may display information relevant to a user because it detected the proximity of that user.
  • Context-triggered actions: these are IF-THEN style rules triggered by contextual information. For example, a watchdog application will execute arbitrary UNIX shell commands based on rules such as "if coffee is made then play the rooster.au sound". A contextual reminder program raises reminders when a contextual rule is satisfied, like "the next time I see marvin" or "the next time I'm in the library".

Relevance

Flaws

• A host of difficulties exist with this framework, such as privacy issues, authentication and security, deciding on the granularity of context changes to recognize or act upon, UI issues for weighting proximity information, networking and OS support for this distribution system spanning many different networking media and computing hosts, etc. Most of these issues are raised but not addressed in the paper.