The document proposes an EEG-based method to evaluate how users perceive the semantic distance of icons. Three experiments were conducted: 1) a function-icon matching task found that close icons attract more attention early on and are distinguished from far icons later; 2) an icon selection task under sliding found novelty and close semantic distance help target recognition, especially fast speeds; 3) a grid selection task found close icons make good targets but poor surroundings, and surrounding effects increase with grid size. The EEG-based method provides a powerful complementary tool to behavioral measures for comprehensively evaluating icons.
Connector Corner: Accelerate revenue generation using UiPath API-centric busi...
CHI 2016: An EEG-based Approach for Evaluating Graphic Icons from the Perspective of Semantic Distance
1. An EEG-based Approach for Evaluating
Graphic Icons from the Perspective of
Semantic Distance
Fu-Yin Cherng, Wen-Chieh Lin, Jung-Tai King, Yi-Chen Lee
National Chiao Tung University, Taiwan
8. Key indication of good icons
[cf. Setlur et al., 2014; Warnock et al., 2013]
Measured by behavior and self-report methods
Effectiveness of conveying information
5
Close
Print
～
～
Far
Semantic Distance
9. [cf. Huang et al., 2015]
Complicated cognitive states and difficult to determine
semantic distance by behavioral measures alone.
Print?
6
T?
10. Complicated cognitive states and difficult to determine
semantic distance by behavioral measures alone.
6
Print?
T?
[cf. Huang et al., 2015]
Use physiological indicators to measure
and analyze cognitive stages.
12. Directly related to cognitive events and states
Used in evaluation and usability testing.
[cf. Chi et al., 2014; Lee et al., 2014]
7
Electroencephalography (EEG) based method
13. [cf. Chi et al., 2014; Lee et al., 2014]
Directly related to cognitive events and states
Used in evaluation and usability testing.
Electroencephalography (EEG) based method
7
EEG-based method is a potentially powerful
tool for evaluating icons.
14. Research Goal
8
Propose EEG-based method to evaluate
human perception of icons, focus on how
users perceive semantic distance of icons.
17. Collection of
Icons
6 functions: Calendar, Crop,
Keyboard, Menu, Print, Setting
70 icons in gray tone
[cf. ICONFINDER; FLATICON; Google Images]
11
18. 50 participants (24 females)
Classify
Semantic Distance
Not Closely
Related
Very Strongly
Related
[cf. Isherwood et al., 2007; Mcdougall et al., 1999]
12
Far iconsClose icons
26. Potential(μV)
Time (ms)
-8
-4
0
4
8
0-100 100 200 300 400 500
Close Far Mismatch
Close Icons attract more attention than far
icons in early cognitive stage.
17
Experiment 1 | Result
Close
Selective Attention
N1
Early Cognitive Stage
27. Potential(μV)
Time (ms)
-8
-4
0
4
8
0-100 100 200 300 400 500
Close Far Mismatch
17
Experiment 1 | Result
0.7
Close
Reaction time (sec)
Far Mismatch
0.9 0.8
Close
Close icon can attract more attention, thereby
shortening reaction time.
N1
28. 18
Experiment 1 | ResultPotential(μV)
600
-8
-4
0
4
8
Time (ms)
0-100 100 200 300 400 500
Close Far Mismatch
30. Potential(μV)
600
-8
-4
0
4
8
Time (ms)
0-100 100 200 300 400 500
Close Far Mismatch
Semantic distance level is distinguished in
later cognitive stage.
20
Experiment 1 | Result
Semantic Incongruence
N400
Later Cognitive Stage
31. Potential(μV)
600
-8
-4
0
4
8
Time (ms)
0-100 100 200 300 400 500
Close Far Mismatch
21
Experiment 1 | Result
N400
Close
Error Rate (%)
Mismatch
0.7
19.9
5.6
Far
Close
Mismatch
Opposite groups of semantic incongruence
reduce error rate.
32. Potential(μV)
600
-8
-4
0
4
8
Time (ms)
0-100 100 200 300 400 500
Close Far Mismatch
Vague semantic incongruence increases
error rate.
22
Experiment 1 | Result
Close
Error Rate (%)
Far Mismatch
0.7
19.9
5.6
N400Far
33. Participants’ behaviors provided basic findings,
EEG results revealed causes of behaviors and
performance in different cognitive stages.
http://www.userzoom.com/wp-content/uploads/2015/04/usability-lab.jpg
42. 26
Experiment 2 | Result
Potential(μV)
600
10
5
0
-5 N1
Fast & Close Fast & Far
Slow & Close Slow & Far
0-100 100 200 300 400 500
Time (ms)
15
600
Target icon shown
43. 27
Far target icons are easily ignored in fast
speed.
Experiment 2 | Result
Potential(μV)
600
10
5
0
-5 N1
Fast & Close Fast & Far
Slow & Close Slow & Far
0-100 100 200 300 400 500
Time (ms)
15
600
Selective Attention
N1
Early Cognitive StageFast & Far
44. Potential(μV)
600
10
5
0
-5 N1
Fast & Close Fast & Far
Slow & Close Slow & Far
0-100 100 200 300 400 500
Time (ms)
15
600
28
Experiment 2 | Result
Novelty in a Series
of Information
N2
45. Potential(μV)
600
10
5
0
-5 N1
Fast & Close Fast & Far
Slow & Close Slow & Far
0-100 100 200 300 400 500
Time (ms)
15
600
29
Close target icons are easily recognized in
fast speed.
Experiment 2 | Result
Novelty in a Series
of Information
N2
Fast & Close
46. Potential(μV)
600
10
5
0
-5 N1
Fast & Close Fast & Far
Slow & Close Slow & Far
0-100 100 200 300 400 500
Time (ms)
15
600
30
Close target icons are easily updated to
working memory.
Experiment 2 | Result
Working Memory
Updating
P3b Far
Close
47. Novelty and close semantic distance of
target icons are important, especially when
searching in fast speed.
http://oemsolutions.agameautotrader.com/wp-content/uploads/2015/01/185649173.jpg
59. Surroundings: Close
Small Gird Size
Big Gird Size
Surroundings: Far
Make trade-offs between reaction time and
error rate based on screen size of applications.
http://www.smartwatchandroid.com/wp-content/uploads/2013/10/sony_smartwatch2-1.jpg
http://files.technobezz.com/files/uploads/2015/05/ipad.jpg
64. • Identify perceptual effects of icons
• Provide more refined method for evaluating icons
• Demonstrate how findings from EEG enrich icon
usability testing.
Fu-Yin Cherng | fufu22710@gmail.com
Questions?
An EEG-based Approach for Evaluating Graphic
Icons from the Perspective of Semantic Distance
Wen-Chieh Lin | wclin@cs.nctu.edu.tw
National Chiao Tung University, Taiwan
