2. Recurring theme with TUIs
• Lots of potential benefits for learning:
• embodied cognition
• multiple external representations, dyna-linking
• improved collaboration
• more in Marshall, “Do tangible interfaces really
enhance learning?” (2007)
• But no empirical evidence of learning
gain (most studies rather report on
engagement)
2
Thursday, November 15, 12
3. Our hypothesis
“Tangibles are neither activators nor inhibitors of
learning; rather, what makes a difference for the
learning performance is how the learning
activities make use of them.”
3
Thursday, November 15, 12
4. Illustration: varying the type
of feedback
• Feedback: in the top 5 influences on learning
(Hattie, 2007)
• Immediate vs. delayed feedback: old debate in
education
• Process and task level feedback:
• process: the main process used to perform the task
• task: how well tasks are understood / performed
• General learning theory; does it hold for TUIs?
4
Thursday, November 15, 12
5. Feedback in TUIs
• Task level
• Process level: if immediate, similar to dyna-
linking
• one of the often cited advantages of tangibles
• but, beware of the “manipulation temptation” (e.g. Do-
Lenh 2012)
• We study the impact of dyna-linking on the
learning performance in our TUI
5
Thursday, November 15, 12
6. Our research questions
1. Is dyna-linking beneficial for learning in our TUI?
2. What is the impact of giving immediate feedback
vs. delayed feedback at the process level?
6
Thursday, November 15, 12
7. Experiment
• Participants:
• 56 carpenter apprentices aged between 16 and 21
• 8 exercises with each of the 3 blocks
• completed the task in pair
• Conditions:
• coupling: immediate process-level feedback
• no-coupling: no process-level feedback
• Learning gain:
• 5-10 minutes pre-test and post-test
• on paper, matching pairs of <side,face> views with top
view
7
Thursday, November 15, 12
10. Learning gain
• Positive overall learning gain
(post-test score – pre-test score)
• When conditions are separated, learning gain difference is only
significant for no-coupling
10
Thursday, November 15, 12
11. Answer types analysis:
pre-test / post-test
Ratio between pre-test and post-test answer types
¬x¬y
Fewer mistakes
on the y-axis for
¬xy
no-coupling
x¬y
xy
0 0.375 0.75 1.125 1.5 1.875 2.25
11
Coupling No coupling
Thursday, November 15, 12
12. Orthographic projections
z-axis
Face view Side view
Top view
Side view
x-axis y-axis
Top view y-axis
Face view
12
Thursday, November 15, 12
14. Link between the treatment
and the post-test scores
• Correlation between the treatment performance
and the post-test:
• yes for no-coupling (r=0.65, t[13]=3.15, p<.001)
• no for coupling (r=0.35, t[12]=1.28, p>.05).
• No correlation between pre-test and treatment
performance
• This means that the condition had an effect on
the learning performance
14
Thursday, November 15, 12
15. Behavioral differences
during treatment: time
• Coupling faster: 67 vs. 88 sec. per question in
average
• No-coupling waited longer before first move
(14 vs 8 sec.)
• No-coupling: movement speed is higher
15
Thursday, November 15, 12
16. Diversity of paths
zone of correct
answer
no-coupling
coupling
starting point
16
Thursday, November 15, 12
17. Our research questions were
1. Is dyna-linking beneficial for learning in our
TUI?
2. What is the impact of giving immediate
feedback at the process level vs. delayed
feedback?
17
Thursday, November 15, 12
18. Summary of results
• Feedback differences at process level led to
impact at both process and task levels:
• process: dyna-linking encouraged participants to dive
into action
• task: no-coupling led to a better learning
• Many more details in the paper!
18
Thursday, November 15, 12
19. Conclusions and outlook
• Tangibles can improve learning, but...
• ... tangibles by themselves don’t guarantee
learning
• When designing tangibles, important to know
the influence of the level of feedback and how
it is given
• Explore other factors/results from the learning
sciences and how they can be applied to the
design of TUIs
19
Thursday, November 15, 12
21. Our hypothesis
“In the case of spatial reasoning, tangibles can
actually help learn better, because they allow
the learner to make a link between the 3D
physical object and its virtual 2D
representation(s).”
21
Thursday, November 15, 12
22. Our research questions
1. Does it make a difference to have a
tangible object instead of a 3D
representation?
2. What is the impact on learning?
22
Thursday, November 15, 12
23. Task
Find some edges in all 3 views
23
Thursday, November 15, 12
25. Experiment
• Participants:
• 46 carpenter apprentices aged between 16 and 21
• 12 questions in total
• completed the task in pair
• Learning gain:
• 5-10 minutes pre-test and post-test on paper
• Conditions:
• token: the tangible object is a dummy object
that does not resemble the virtual one (10 pairs)
• block: the tangible object has the same
shape as the virtual one (13 pairs)
25
Thursday, November 15, 12
27. Learning gain
• Significant
0.4
Z−score diff. between pre− and post−tests
improvement
overall (+20.8%
absolute gain)
• No significant diff. 0.2
●
between the two
0.0
conditions
• Improvement in
−0.2
block condition is a ●
bit larger (+22.8%
vs. +18.3%)
−0.4
n=13 n=10
block token
Condition
27
Thursday, November 15, 12
28. Performance during treatment
During treatment,
students in the ●
10
block condition
Questions answered correctly
performed
significantly better
9
●
8
7
n=13 n=10
block token
Condition
28
Thursday, November 15, 12
29. Specificity of the side view
15
15
15
# mistakes
Mistakes on elevation
Mistakes on side
Mistakes on plan
10
10
10
●
5
5
5
● ●
●
● ●
0
0
0
n=13 n=10 n=13 n=10 n=13 n=10
block token block token block token
Condition Condition Condition
front top side
Having the block helps especially with the side
view, i.e. the most difficult one
29
Thursday, November 15, 12
30. Orthographic projections
z-axis
Face view Side view
Top view
Side view
x-axis y-axis
Top view y-axis
Face view
30
Thursday, November 15, 12
32. Impact of the degree of symmetry
3.0
• No effect of the View
2.8
view Top
Front
Number of edges found correctly
• More symmetry
2.6
Side
planes leads to
2.4
lower performance
2.2
• Not on this graph,
but no interaction
2.0
effect between the
1.8
condition and the
degree of symmetry
1.6
0 1 3
Degree of symmetry
32
Thursday, November 15, 12
33. Summary and conclusion
• Overall positive learning gain.
• Performance during treatment was higher in the block
condition, but the learning gain is not significantly
different.
• The tangible interface was able to help in particular on
the more difficult aspects of a question (side view).
• A higher degree of symmetry leads to a harder 3D
orientation.
• The design aspects of the TUI and can impact the
learning performance
33
Thursday, November 15, 12