Transcript Project Slides - Personal Web Pages

Projects
March 29, 2006
Project Requirements
• Think Aloud
– At least two people
• OR Difficulty Factors Assessment
– Ideally >25 (at least one class), but negotiable with
instructors
– OR Educational Data Mining
• Pseudo-tutor & design rationale
• Skill matrix & problem set
• Cognitive model & cognitive tutor
– Model tracing, bug messages
– Hint messages
Project Scoping Proposal
• 2 paragraphs
• Project title, participants, areas of relative
expertise
• Describe the subject domain you intend to build a
tutor for
• Provide an example of a task that students might
perform in your tutor
• Provide at least one hypothesis about what is hard
for students in this domain
Project Ideas
• Suggestion: Pick a college level topic to
make data collection easier
• Some ideas:
–
–
–
–
–
Logic proof tutor
Lower-level logic rule tutor
Recursion, Induction
Combinatorics
Intro programming
• Other ideas?
Logic tutor components
• Deep Thought graphical proof solver
• Corresponding Text-based proof
• Feedback:
– “I’m stuck” help button
– Roll-over descriptions
– Automatically generated
• Incorrect actions
• Off-path to solution
• Too long before input
Logic tutor continued
• Adaptive interface with levels
– Beginner
• Tutoring on basic rules and how they work
• Tutoring on interface?
• Simpler problems
– Intermediate
• Ask for reasons for every move
• More difficult problems
– Advanced
• Don’t have to justify every step
• Planning interface?
• Most difficult problems
Logic tutor continued
• Intelligence required:
–
–
–
–
Tracing the student’s behavior
Expert model that can execute proof
Be able to suggest next step
Be able to detect student’s “plan”
Architecture
• Deep Thought is in Java
• JESS – rule production system – getting
license, Sandia Labs
• Drools – open source
• NovaNET Proofs tutorial
– TUTOR
– Dr. Barnes will write pseudo-code for this
•
•
•
•
Build textual formal proof window
Rearrange the interface – move buttons
Add Save feature, Authentication in our DB
Database – JDBC in code,
– RMI (remote method invocation)
– Oracle database, have sequel server – both at UNCC
– Check at NCSU
• Create problem set –
– Tiffany, 25 & 15, order changes
• Interface help: movie on how to use?
• Write capabilities of each – better understanding
of features of Deep Thought
• Load old data into database
• Storyboard the levels
• Production rule system, with correct and
incorrect rules marked
• Hint button
– Multiple levels of hints
•
•
•
•
Ask a question: did you think about …?
Combine the highlighted lines for the next step
Possibly highlight steps and rule
Do it for them
• Help interface – movie or text based
• Feedback
– Mark when something is wrong
•
•
•
•
Based on student model?
More detail at beginner level
At advanced level
Example: click on two items and a rule, if rule
doesn’t apply can say “Rule does not apply”
– If do it again, refer to intro tutorial
– Indicate when a solution is not on path
•
•
•
•
At beginner level, immediately
Intermediate, maybe after 2 steps
Advanced – maybe after 4 steps
Might ask for feedback about plan?
• NovaNET: bit gets set that correct answer
attained
• Print option – May be hard, not important
• Load partial or completed problem option
• Keep track of student model/score/skills –
reflect in score, or profile that students can
view.
• Keep track of and show time
• Add functionality to show how a rule
actually got applied
•
•
•
•
•
•
Think aloud with 2 students
Interview experts – Dr. Barnes & Dr. Croy
Data mining of both programs
Dr. Barnes will get her data ready
John will get Deep Thought data ready
Invite Dr. Croy to class for an interview