Skip to main content

Out of Context

Computer Systems that Adapt to and Learn from Context​

Introduction​

This paper by Lieberman and Selker addresses the limitations of traditional computer systems that rely solely on explicit user input to produce outputs. It advocates for systems that are context-aware—meaning they can adapt based on a variety of environmental, user-specific, and historical factors. This shift toward contextual sensitivity is seen as crucial for advancing artificial intelligence (AI) and improving human-computer interaction (HCI).

Target​

The concepts in this paper apply broadly to both desktop and embedded interfaces and aim to influence system-level design principles for a wide range of applications, including embedded computing.

Key Insights​

  • Definition and Importance of Context: Context is defined as any information about the user's environment, preferences, or historical interactions that influence system behavior. Recognizing context is vital for systems to act appropriately and reduce user frustration.
  • Limitations of the "Black Box" Model: Traditional systems are context-independent, which limits their responsiveness and flexibility. The paper suggests expanding system boundaries to consider context as an implicit input.
  • Abstraction vs. Context Sensitivity: There is a trade-off between maintaining high levels of abstraction (ignoring context) and adapting to specific contexts, which often enhances user satisfaction and system usefulness.
  • Context in System Models:
    • User Model: Tailors system responses based on user preferences and behaviors.
    • Task Model: Adjusts system actions based on the current task.
    • System Model: Includes data about the computational environment, which enables context-aware responses.

Supporting Data​

  • Examples from AI and Agent Systems: The authors discuss systems like COACH, an adaptive help system that personalizes guidance based on user experience, and Letizia, a web browsing assistant that predicts user interests based on browsing history.
  • Embedded Context-Sensitive Devices: Examples include "Talking Trivets" (sensors in kitchen tools) and context-aware furniture that responds to user presence and schedules, illustrating the potential for everyday objects to become adaptive.

Other Insights​

  • Role of User Feedback: Effective context-sensitive systems should incorporate user critiques and modify behavior over time, leading to more accurate responses based on user rejection or correction of system actions.
  • Implications for Ambient Interfaces: Ambient or subtle interfaces could help reduce the cognitive load on users by offering context-sensitive hints without explicit interaction, such as environmental cues.

Practical Applications​

  • Design Context-Aware Agents: Develop software agents that can adapt based on user history and inferred preferences, such as recommending relevant options based on previous choices.
  • Incorporate Context in Embedded Devices: Enable embedded devices with sensors to dynamically adjust their behavior, like adjusting brightness in a room based on ambient light.
  • User Model Integration: Systems should consider factors like expertise level, past actions, and preferences to deliver customized responses and avoid generic behavior.
  • Enhance System Feedback Loops: Design systems with feedback mechanisms to update and refine context sensitivity, which allows the system to become more intuitive over time.

Reference​

Lieberman, H., & Selker, T. (2000). Out of Context: Computer Systems that Adapt to, and Learn from, Context. IBM Systems Journal, 39(3&4), 617–632.