Context in Artificial Intelligence

  • Guha & Lenat (1991): Contexts are microtheories.

  • Giunchiglia (1992) : making reasoning local

  • McCarthy (1993) : extension of Logic by ist(C, P) formula

  • Sowa (1992, 2000): extension of conceptual graphs

  • Experts state knowledge in a specific context as a triplet (problem, context, solution), but the knowledge engineer acquires the knowledge, not the specific context as a couple (problem, solution) for generalization purpose .

  • In learning

    - Concept meaning depends on some implicit context
    - Changes in a context cause changes in concepts and their relationships (e.g. the distance between two concepts is context dependent)

    - Context can determine which attributes are predictive at any point in time

    Thus, in AI, context is used for: - either theoretical purposes.
    - or practical needs

    Typically no modeling, only representation.