Robert Bixler amp Sidney DMello rbixlerndedu University of Notre Dame July 10 2013 mind wandering indicates waning attention occurs frequently 2040 of the time decreases performance ID: 440514
Download Presentation The PPT/PDF document "Toward Fully Automated Person-Independen..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
Toward Fully Automated Person-Independent Detection of Mind Wandering
Robert
Bixler
& Sidney
D’Mello
rbixler@nd.edu
University of Notre Dame
July 10, 2013Slide2Slide3
mind wandering
indicates waning attention
occurs frequently
20-40% of the time
decreases performancecomprehensionmemorySlide4
solutions
proactive
mindfulness training
Mrazek
(2013)tailoring learning environmentKopp, Bixler
,
D’Mello
(2014)
reactive
mind wandering detectionSlide5
our goal is to detect mind wanderingSlide6
related w
ork
–
attention
Attention and Selection in Online Choice TasksNavalpakkam
et al. (2012)
Multi-mode Saliency Dynamics Model for Analyzing Gaze and Attention
Yonetani
, Kawashima, and Matsuyama (2012
)
distinct from mind wanderingSlide7
mind w
andering
d
etection
neural activityphysiology
acoustic/prosodic
eye movementsSlide8
neural activity
Experience Sampling During fMRI Reveals Default Network and Executive System Contributions to Mind Wandering
Christoff
et al. (2009)Slide9
physiology
Automated Physiological-Based Detection of Mind Wandering during Learning
Blanchard,
Bixler
,
D’Mello
(2014)Slide10
acoustic-
p
rosodic
In the Zone: Towards Detecting Student Zoning Out Using Supervised Machine Learning
Drummond and
Litman
(2010)Slide11
eye m
ovements
m
indless
r
eading
m
indful
r
eadingSlide12
research questions
can mind wandering be detected from eye gaze data?
which features are most useful for detecting mind wandering?Slide13
4 texts on research methodsself-paced page-by-page
30-40 minutes
difficulty and value
auditory probes
9 per text inserted psuedorandomly (4-12s)
data collection
type of report
yes
no
total
end-of-page
209
651
860
within-page
1278
2839
4117
total
1487
3490
4977
t
obii
tx300Slide14
compute fixations OGAMA (Open Gaze and Mouse Analyzer)
(
Voßkühler
et al. 2008)
compute features
build supervised machine learning models
data analysisSlide15
globallocal
context
featuresSlide16
global features
eye movements
fixation duration
saccade duration
saccade lengthfixation dispersionreading depthfixation/saccade ratioSlide17
local features
reading patterns
word length
hypernym
depthnumber of synonymsfrequencyfixation type
regression
first pass
single
gaze
no wordSlide18
context features
positional timing
since session start
since text start
since page startprevious page timesaverageprevious page to average ratio
task
difficulty
valueSlide19
supervised machine learning
parameters
window size (4, 8, or 12)
minimum number of fixations (5, 1/s, 2/s, or 3/s)
outlier treatment (trimmed, winsorized, none)feature type (global, local, context, combined)downsampling
feature selection
classifiers (20 standard from
weka
)
leave-several-subjects-out cross validation (66:34 split)Slide20
1. can mind wandering be detected using eye gaze data? Slide21
1. can mind wandering be detected using eye gaze data? Slide22
1. can mind wandering be detected using eye gaze data?
confusion matrices
end-of-page within-page
actual response
classified response
prior
yes
no
yes
.54
.46
.23
no
.23
.77
.77
actual response
classified response
prior
yes
no
yes
.61
.39
.36
no
.42
.58
.64Slide23
2. which features are most useful for detecting mind wandering?Slide24
2. which features are most useful for detecting mind wandering?
rank
end-of-page
within-page
1
previous value
saccade
length max
2
previous difficulty
saccade
length median
3
difficulty
fixation
duration ratio
4
value
saccade length range
5
saccade length max
saccade length mean
6
saccade length range
saccade length skew
7
page number
fixation duration median
8
saccade length
sd
fixation duration mean
9
saccade
length mean
saccade duration mean
10
saccade length skew
saccade duration minSlide25
summary
mind wandering detection is possible
kappas
of .
28 to .17end-of-page models performed betterglobal features were best
exception: context features highest ranked for end-of-pageSlide26
enhanced feature set
global
pupil diameter
blink frequency
saccade anglelocalcross-line saccadesend-of-clause fixationsSlide27
enhanced feature setSlide28
predictive validity
mw
rate
post
knowledge
transfer
learning
end-of-page
predicted
-.556
-.415
actual (model)
-.248
-.266
actual
(all data)
-.239
-.207
within-page
predicted
-.496
-.431
actual (model)
-.095
-.090
actual (all data)
-.255
-.207Slide29
self-caught mind wanderingSlide30
what does mind wandering look like?
saccades
slower
shorter
more frequent blinkslarger pupil diametersSlide31
limitations
eye tracker cost
population validity
self-report
classification accuracySlide32
future work
multiple modalities
different types of mind wandering
mind wandering interventionSlide33
acknowledgements
Blair Lehman
Art
Graesser
Jennifer NealeNigel BoschCaitlin MillsSlide34
questions
?