Slides for my talk "Misconceptions in Visual Algorithm Simulation Revisited: On UI’s Effect on Student Performance, Attitudes, and Misconceptions" from Learning and Teaching in Computing and Engineering (LaTiCE) 2013.
The paper can be found from: http://doi.ieeecomputersociety.org/10.1109/LaTiCE.2013.35
Misconceptions in Visual Algorithm Simulation Revisited
1. Misconceptions in Visual Algorithm
Simulation Revisited:!
On UI’s Effect on Student Performance, Attitudes, and
Misconceptions
LaTiCE 2013, March 22
Ville Karavirta, Ari Korhonen, Otto Seppälä
Aalto University, Finland
3. Algorithm Visualization & Simulation
• Many, many wonderful AV systems !
– Goal to help students learn algorithms
• Most implemented in Java "
• Algorithm simulation: activate!
students and make them simulate!
the algorithm
6. Binary Heap
• Binary tree with heap-property:
– MaxHeap: father greater than child
7. BuildHeap Misconceptions
• BuildHeap algorithm
– “Turns an array into a heap”
• Misconception: “non-viable mental models that
students themselves assume to be correct“
• Previous research by Seppälä et al. in 2005
– Identified several build-heap misconceptions
10. Research Setup
• CS Majors undergraduate course
• Priority Queues exercises replaced
– Binary Heap Insert
– Binary Heap Delete
– Build-heap
– Heapsort
• Rest of the exercises still used TRAKLA2
• Feedback Questionnaire
11. Research Questions?
1. Do students prefer a more simplified UI and
interaction?
2. Do students have the same misconceptions
as in the earlier studies?
3. Do we catch the student misconceptions?
13. Identifying the Misconceptions
• Manually implemented incorrect versions of the
build-heap algorithm
• Graded the incorrect student answers with
those implementations
• If an incorrect algorithm gave full score, label the
answer to have that misconception
16. Student Attitudes
• “Students liked it” !
• Visual appearance got most mentions
– Positive and Negative
• Other common themes: smooth animation,
explanations in model answers
• Some technical issues
• 56% would like to solve the exercises on mobile
devices
18. Possible Remaining Misconceptions
Answers that would have gotten full grade with
both the correct and a misconceived algorithm
Misconception
Answers
Answers %
Wrong-Duplicate
98
89.9%
Left-to-Right
29
26.6%
Heapify-with-Father
41
37.6%
Delayed-Recursion
49
45.0%
Other
7
6.4%
Smallest-Instantly-Up
1
0.9%
No Possible Misconceptions
8
7.3%
19. Dealing with Remaining Misconceptions
• Focus more on input data randomization
– Two potential solutions
1. Select from a predefined set of input data
2. Ensure that no known misconception gives full grade
for the input data
• Improved for 2013 course, using solution 2
– Data collected, not analyzed yet
20. (Automatic) Recognition of Misconceptions
• Testing for a misconception requires knowledge
about the misconceptions
• Approaches
– Code mutation [Seppälä 2006]
– Give same input to many students, find repeating
solution sequences
– “Normalize” the inputs given, then find repeating solution
sequences
• Manual verification/labeling of the misconceptions
22. Conclusions
• New UI had no significant effect on
– student results
– student misconceptions
• Real care must be used when randomizing
input data
• Next step: automatic recognition of
misconceptions