BOZ: An Automated Graphic Design and Presentation Tool

1
BOZ An Automated Graphic Design and
Presentation Tool
Summarized from Casner, S. M., A Task-Analytic
Approach to the Automated Design of Graphic
Presentations. ACM Transactions on Graphics,
Vol. 10, No. 2 April 1991, p. 111 151. Casner,
S. M, and Larkin, J. H., Cognitive Efficiency
Considerations for Good Graphic Design. In
Proceedings of the Eleventh Annual Conference of
the Cognitive Science Society, Ann Arbor, MI,
August 1989, 11 pp.
2
Author Biography

• Stephen Casner
• Education
• BS, MS - Computer Sciences
• PhD - Intelligent Systems, University of
  Pittsburg, 1990.
• Current Work
• NASA Ames Research Center Aviation Training
  Research
• Cockpit automation and training
• Avid pilot holding many certifications and ratings

3
Scope of Research

• Casner (1991)
• Focus on task to be performed using graphs
• Theory and design elements behind BOZ
• Working example
• Advantages and limitations
• Casner and Larkin (1989)
• Summary of BOZ design and operation
• Empirical study

4
Questions

• What could be done to enhance BOZ to produce
  unique or innovative graphical displays?
• Can every useful logical operator be expressed as
  a perceptual operator?
• Are there tasks that cannot be expressed well as
  perceptual procedures?
• Does BOZ introduce unacceptable limitations on
  graphical presentations beyond those identified
  by the author?
• What advantages does BOZ afford over prior
  empirically-based graphic design?
• Could using the University of Pittsburg employees
  influence test results?
• Is the study population really large enough to
  tell us anything about the use of BOZ in the real
  world?
• Does BOZ address possible need for memory recall
  of facts and figures?

5
BOZ Basic Premises

• What we can hope to achieve in an automated
  graphic design tool is a codified set of
  operational design principles that perform
  satisfactorily across interesting tasks and
  graphics.
• Graphs can expedite human information processing
  by substituting perceptual inferences for
• Computation
• Searching
• Effective presentation cannot be separated from
  task to be performed
• Focus on delineating efficient and effective
  perceptual procedures
• Use formal a framework to identify/codify
  perceptual procedures

6
Five BOZ Components

• Logical Task Description Language
• Perceptual Operator Substitution
• Perceptual Data Structuring
• Perceptual Operator Selection
• Rendering

7
Logical Task Description Language

• Allows encoding and description of task steps
  Compose sentential data structure and program
  from Larkin and Simon (LS)
• Read by BOZ as input
• Consists of logical procedures that act on
  logical facts
• Domain set
• Logical Operators (LOPs)
• Main procedure body

8
Logical Task Description Language

• Domain Set
• Types of information/data describing system
• Three types
• Ordinal (list)
• Nominal (count)
• Quantitative (values)
• LOPs
• Functions used by logical procedure to search,
  manipulate and compute logical facts
• Two types
• Search
• Computation
• Procedure Body
• Task-oriented function that executes a series of
  LOPs using standard program control constructs
  (loops, logical, etc.)
• Maintains logical facts

9
Perceptual Operator Substitution

• Identify perceptual operators (POPs) that may
  potentially replace LOPs
• Maintain informational and computational
  equivalence between LOPs and POPs LS
• POP
• Analogous to LOPs, but describe perceptual
  inferences or visual search using graphic
  information
• Associated with a graphical representation object
• Two types
• Search
• computation
• Maintained in a catalog and organized by
  equivalent classes
• Compute same type of information i.e.,
  distances, widths, etc.

10
Perceptual Operator Substitution

• POP (cont.)
• Based on graphic primitives
• Height
• Width
• Color
• Etc.
• Substitution is based on POP equivalence to LOP
• LOP is compared to POPs on an equivalence class
  basis
• If LOP is equivalent to on POP in the class, it
  is considered equivalent to all other POPs in
  that same class

11
Perceptual Data Structuring

• Determines optimal grouping and distribution of
  information within a graphic
• Location and diagrammatic data structure
  philosophy from LS
• Examines domain sets manipulated by LOPs and
  determines
• Information to portray in a graphic
• Spatial locality of operators
• Partitioning and presentation of non-shared
  information

12
Perceptual Data Structuring

• Operator Vectors
• Vectors provide explicit identification of domain
  sets or feature space manipulated by an LOP
• Two types
• Search (object, attribute)
• Computation (domain sets manipulated)
• Search vectors describe how information can be
  related and grouped into graphical objects
• Search vector relationship types
• Conjoint Vectors that contain common objects
• Parallel Vectors that are related through a
  common attribute manipulated by a computation
  vector
• Orthogonal Vector A is orthogonal to Vector B
  if it contains the save attribute as Vector B
• Disjoint Vectors that contain no common object
  or attribute

13
Perceptual Operator Selection

• Choose single best POP to replace a LOP
• Selection based on
• efficiency and accuracy
• representational power of the graphic primitive
• ability to combine with other operators to
  generate coherent graphical object
• BOZ relies on APT model for quantifying
  efficiency, accuracy and representational power
• Ability to combine operators is dependent on
  composition rules for graphical presentation
  objects associated with POP

14
Rendering

• Translates logical fact into graphical facts
• Provides physical output for graphic user
• Provides for interaction with user
• Changes to logical facts
• Changes to graphical objects
• Relationship/mapping is done via the perceptual
  data structure
• Ensures that graphic as presented matches the
  logical fact and makes sense from a visual
  perspective

15
Analysis

• BOZ goals met
• Efficient substitution of perceptual inferences
  in place of logical inferences
• Reduced the number items considered when
  searching for needed information
• Limitations
• No algorithmic strategy that always chooses the
  best POPs
• Generalization may not be possible
• Level of skill
• Practice
• Task
• Particular presentation
• Age
• Culture
• Social situation

16
Analysis

• Limitations (cont.)
• No algorithmic strategy that always chooses the
  best POPs
• Problem domain not well-defined impedes
  automated selection of appropriate POPs
• POP selection is made independent of context or
  combinations of POPs that may be more efficient
• Rendering component is insensitive to
  domain-specific conventions
• Does not make use of real-world conventions and
  artifacts when rendering graphics
• Choice of font type and graphics color are not
  addressed in BOZ

17
Empirical Study Results

• Test Configuration
• Four graphic displays of airline reservation
  system replacing LOPs with POPs
• LOPs include
• findFlight(origin city1, destination city2)
• checkAvailability(flight)
• checkLayover(flightA, flightB)
• checkCost(flightA, FlightB)
• Displays include
• Display 1 (classic table) findFlight,
  readAvailability(flight), subtractTimes(flight1,
  flight2), addCosts(flight1, flight2)
• Display 2 (horizontal distance encodes times)
  All POPs from Display 1 plus findHorzDistance
  (replaces checkLayover LOP)

18
Empirical Study Results

• Displays include (cont.)
• Display 3 (shading encodes availability) All
  POPs from Display 2 plus judgeShaded (replaces
  checkAvailability LOP)
• Display 4 (height encodes cost) All POPs from
  Display 3 plus judgeHeights (replaces checkCost
  LOP)
• Methods
• Participants University of Pittsburg employees
  (eight)
• Materials 40 test cases, composed of ten
  instances of each of display
• Apparatus 9x12 screen images of each test case
• Procedure Perform reservation task forty times,
  ten times using each display
• Presented in eight different orders
• All visual operators explained
• Taught rowSearch procedure but allowed to use
  any strategy
• Measured time needed to interpret each display

19
Empirical Study Results

• Hypotheses
• Global Efficiency
• Cognitive efficiency should be linearly ordered
  display 1 through 4 should have increasing
  efficiency, respectively
• Decrease in operator times
• POPs should require less time to interpret than
  LOPs increasing use of POPs in display 1 through
  4 respectively should likewise result in
  decreased time to complete task
• Results
• Global Efficiency
• Mean response times (seconds)
• Display 1 (table) 19.3 (s.d. 8.4)
• Display 2 (horizontal distance encodes time)
  10.1 (s.d. 4.7)
• Display 3 (shading encodes availability) 7.2
  (s.d. 2.7)
• Display 4 ( height encodes cost) 7.4 (s.d. 2.4)

20
Empirical Study Results

• Results
• Decrease in Operator Times
• Search step requires 335 35 ms
• findHorzDistance POP is 2 0.25 s faster than
  subtractTimes LOP
• JudgeHeights POP is 100 300 ms slower the
  addCosts LOP
• Differences in Display 1 through 3 explained by
  efficiency of POPs
• Lack of difference in Display 3 and 4 explained
  by need to integrate visual information from two
  different locations in the judgeHeights POP
• The Bottom Line
• Operator substitution and use of visual cues
  helps to reduce search time restructuring of
  search did not seem to help
• BOZ can be helpful in reducing the human
  processing needed to perform a cognitive task.

21
Last Words

• Design of Effective Graphical Presentations
• BOZ Takes task-specific logical operators and
  converts them to efficient perceptual operators
  for perceptual processing
• Learning Graphical Conventions and Procedures
• BOZ does not address the impact of learning on
  efficiency of perceptual processing, HOWEVER
• Problem should be minimal since BOZ only uses
  conventional and familiar perceptual operators
• Real-Time Automated Graphical Presentation
• BOZ can be used interactively to automate and
  generate graphical presentations
• Executable Logical and Perceptual Procedures
• Can use BOZ to simulate changes in presentation
  to find most effective graphical presentation

22
Questions

• What could be done to enhance BOZ to produce
  unique or innovative graphical displays?
• Can every useful logical operator be expressed as
  a perceptual operator?
• Are there tasks that cannot be expressed well as
  perceptual procedures?
• Does BOZ introduce unacceptable limitations on
  graphical presentations beyond those identified
  by the author?
• What advantages does BOZ afford over prior
  empirically-based graphic design?
• Could using the University of Pittsburg employees
  influence test results?
• Is the study population really large enough to
  tell us anything about the use of BOZ in the real
  world?
• Does BOZ address possible need for memory recall
  of facts and figures?