Some directions for studying the "objects first" question

1
Some directions for studying the "objects first"
question
2
Approaches to teaching OO in CS 1 and 2

• Graduated exposure with scaffolding (Buck
  Stucki, SIGCSE 2000)
• Complex system that can be extended with class
  definitions (Nevison Wells, ITiCSE 2003)
• Emphasis on early analysis and design, coverage
  precedes procedural constructs tool translates
  class diagrams to code (Alphonce Ventura,
  ITiCSE 2003)
• Conceptual level first with UML code with
  objects first, then internal construct of classes
  (Bennedsen Caspersen, SIGCSE 2004)
• Object-centered design (Adams Frens, SIGCSE
  2003)

3
Buck Stucki

• Design at the beginning is premature
• Process
• Start with code skeletons and internal
  specifications
• Students write code in skeletons
• Incrementally, structure is removed and
  higher-level specifications are offered
• Authors claim Blooms taxonomy indicates an
  initial need to start with low-level constructs
  (e.g. words before sentences before paragraphs)

4
Nevison Wells

• Students need term-long case-studies to motivate
  design and programming
• Process
• Use complex system (e.g. elevator simulation) as
  case study throughout the term
• Students write classes that specify behavior of
  objects in system (e.g. write code that moves an
  elevator)
• Students learn more about system, extend
  functionality
• Cite use of other case studies including the
  Marine Biology Simulation

5
Alphonce Ventura

• Emphasize design over code and syntax
• Use code-generating tools based on design
  diagrams
• Non-trivial examples
• Uses UML and design patterns
• Similar to Model-First approach by Bennedsen
  Caspersen, who note a passing rate of 79
  compared to 52 under previous approach

6
Adams Frems

• Object-centered Design Students need a gradual
  approach to learning OO methodology
• Process (all within CS 1)
• Phase I Objects early---uses objects of
  predefined classes
• Phase II Classes and methods---start with class
  methods, interleave control structures
• Phase III Inheritance
• Anecdotal evidence of office hour visitations
  fewer comments of I dont know where to start

7
Overview of approaches

• All emphasize OO concepts from the beginning
• Complex term-long cases vs. simple examples
• Model-based vs. code-based
• Anecdotal or informal evaluation (if any)
• No detailed principles for motivating approach

8
Using a controlled study

• Method
• Establish objective measures of student learning
• Conduct multiple approaches
• Control for confounding variables statistically
  or through random assignment
• Drawbacks
• Expensive or impractical
• Results may not generalize
• Does not explain why result was achieved

9
Alternate approach

• Identify common misconceptions
• Develop mental models based on misconceptions
• Develop questions for diagnosing misconceptions
• Use mental models to create examples and
  activities that target misconceptions
• Conduct small experiments for assessing
  effectiveness of examples or activities

10
Possible mental models

• Objects are useful for qualifying the method
  there is no state associated with an object.
• A class is written for one object. A state is
  associated with the class-object pair. Additional
  objects belonging to the class dont make sense
  or they all share the same state.
• Each object has its own state. A class is written
  as a template for creating objects and describing
  their behavior.

11
Diagnostic questions

• Create multiple objects belonging to a class,
  each with different states ask students to trace
  code and predict behavior
• Does it make a difference if they need to trace
  through the methods?
• What diagrams do they create for tracing code? Do
  they have a state for each object?
• When writing a class to a specific object, do
  they try to reference the object name in the
  class definition?

12
Teaching examples which ones work?

• Standard input object
• File input objects
• String objects
• Temperature object
• Object holds temperature
• Methods produce values for other units
• MarbleJar object
• Simulates marble jar with draw and add methods
• Turtle graphics objects

13
How do we start?

• Investigate misconceptions
• Protocol Analysis
• K. A. Ericsson and H. A. Simon
• Please think aloud!
• Conceptual inventories
• Frequency profiles of misconceptions/mental
  models
• Mini-experiments

14
Conferences and Journals

• SIGCSE (ACM SIG on computer science education)
• ITiCSE (international conference)
• Computer Science Education (journal)
• International Conference of the Learning Sciences
• Journal of the Learning Sciences
• Cognitive Science (conference and journal)