Keith A. Pray - Professional and Academic Site
About Me
·
·
·
LinkedIn Profile Facebook Profile GoodReads Profile
Professional
Academic
Teaching
                                          
Printer Friendly Version

Intro ] [ Final-Project ]

Critiques ] [ Presentations ] [ Up: Academic ]


AI in Design System Profile
  1. General
    • Domain:
    • System Name:
    • Dates:
    • Researchers:
    • Location:
    • Related Systems:

  2. Knowledge
    • Representation:
      • Is the representation numerical, text-based, mixed?
      • Are numerical formulas used?
      • Are more abstract formulas used?
        indicating sign and/or level of magnitude
      • How redundant is the knowledge?
        How many different representations can the same piece of data or knowledge have in the system?
      • Does the knowledge representation incorporate a high level of cross-indexing / referencing?
    • Input:
      • How is the design problem represented?
      • Are there functional, behavioural or constraint-based aspects to the input?
      • Is there any domain specific knowledge expected as input?
    • Output:
      • Is the output in the form of a design?
      • Is it a partial design?
      • Is it a redesign or re-design?
      • Is more than one solution given?
      • What is the output if the system fails to generate a solution?
      • Can incorrect solutions be given?
      • If so, what other information is provided?
    • Information Contained in System:
      • What domain specific knowledge does the system have?
      • What non-domain specific knowledge does the system have?
      • What domain specific control knowledge does the system have?
      • What non-domain specific control knowledge does the system have?
      • Are previous solutions utilized as a source of information?
      • What knowledge about design is represented explicitly?
    • Interactive:
      • Can the user add information during design problem solving?
      • Does the user supply any control knowledge?
        partial, full, system requested, user option, etc.

  3. Method
    • Theory:
      • Does the system illustrate the effectiveness of a theory of design?
      • Is the theory based on human problem solving?
    • Weak or Strong method:
    • Advantages of method:
    • Weaknesses of method:
    • Hierarchy:
      • Is knowledge structured in a hierarchy?
      • Is control structured in a hierarchy?
    • Search space:
      • Does the system explicitly search a space of possible solutions?
      • How is the search space represented?
      • What control mechanisms are used to guide the search?
      • Can each point in the search space be viewed as a state?
      • What operators are provided to move from state to state?
        Points in a search space and possible states may seem similar. I mean to differentiate between explicit transition between states, like transversal of a directed graph, and other methods by which points are chosen in a space to be considered.
    • Conflict resolution:
      • Is backtracking used to recover from possibly erroneous decisions?
      • Is the information used for control domain specific?
      • What measures are taken to avoid backtracking?
      • How is conflict between agents handled?
      • How is inner-agent conflict handled?
    • Learning:
      • Does the system improve in performance as it is used?
      • Does the system have any knowledge acquisition mechanisms?
      • If so, what is the source of knowledge?
        user, database, file, previous solutions, simulation, model, etc.
      • Does the system repeat the same mistakes?
      • When is learning done?
        before, during, after problem solving
      • Is feedback used to train the system?
        solution evaluation by the system, the user or some other external system
    • Is the method mainly deterministic?
      Are there any random aspects of the system? Will it produce the same results for the same input on multiple runs?
  4. Architecture
    • Overall structure:
      • Is there a list of sub-tasks?
      • Is there a distinct number of steps taken?
    • Components comprising system:
    • Method for each part:
    • Simulation or Modeling components:
    • Other systems used as a component:
  5. Implementation
    • Language:
    • Machine:
    • Complete or Partial:
    • Performance:
      • Accuracy and precision of results
      • How was this measured?
      • Memory consumption during execution
      • Processing power required to make system viable
      • Run times of example design problem solving
      • Comparison with like systems
    • Data Stores Used:
    • User Interface:
      • Are there menus, buttons, graphics, plain text?
      • What information is displayed about the system during run time?
      • What knowledge about the problem and solution progress is displayed?
      • Is this information displayed for status purposes only or is the user required to take some action for the system to succeed?
    • Real World Use:
      • Was the system used outside the research team?
      • Was the system used by domain experts?
      • Was the system used by domain novices?
      • Is the system suitable for use in a commercial design capacity?

by: Keith A. Pray
Last Modified: August 13, 2004 1:15 AM
© 2004 - 1975 Keith A. Pray.
All rights reserved.

Current Theme: 

Kapowee Hosted | Kapow Generated in 0.022 second | XHTML | CSS