Automatically generating personalized user interfaces with Supple

Document Outline

• Automatically generating personalized user interfaces with Supple
  • Introduction
  • Previous research
  • Representing interfaces, devices and users
    • Functional interface specification (I)
    • Device capabilities and constraints (D)
    • Modeling users with traces (T)
  • Interface generation as optimization
    • The admissible heuristic
    • Constraint propagation
    • Variable ordering
    • Value ordering
  • Formulating the cost function
    • Factorization for efficient computation and personalization
    • Optimizing for expected speed of use
      • Computing EMTmanip
      • Computing a lower bound for EMTnav
    • Capturing consistency across interfaces for different devices
      • Relevant widget dissimilarity features
  • Dynamic personalization of automatically generated UIs
    • System-driven automatic adaptation
    • User-driven customization
  • Evaluation
    • Versatility
    • Adapting to long-term usage patterns
    • User-driven customization
    • System performance
      • Variable ordering heuristics and the parallel algorithm
      • Scalability
      • Importance of constraint propagation
    • Performance when optimizing for expected speed of use
    • Model complexity
  • User evaluation
    • Overview of the approach
    • Participants
    • Apparatus
    • Part 1: Eliciting personal models
      • Preference elicitation tasks
      • Ability elicitation tasks
      • Procedure
      • Note on the validity of preference models
    • Part 2: Main experiment
      • Tasks
      • Procedure
      • Generated interfaces
      • Design and analysis
    • Results
      • Adjustment of data
      • Completion times
      • Error rates
      • Subjective results
  • Discussion
  • Conclusion
  • Acknowledgements
  • References