Transcript Mobile Phone Use in a Driving Simulation Task: Differences

Mobile Phone Use in a Driving
Simulation Task: Differences in
Eye Movements
Stacy Balk, Kristin Moore,
Will Spearman, & Jay Steele
The Problem
Each year there are nearly 43,000 traffic
collisions (NHTSA, 2005)
Traffic crashes are responsible for 40
percent of deaths of people aged 15-20
(National Transportation Board, 2005)
Inattention is the most sighted cause for
traffic crashes (NHTSA, 2000)
Background
When driving, & mental workload is
increased (e.g. high traffic, visual clutter,
etc.) drivers are less able to maintain high
situation awareness.
A reduction in situation awareness may
result in a lowered ability to optimally
perform driving tasks (Gugerty 1997).
Background Cont.
In addition to normal aspects of driving,
conversing on mobile phones has been
shown to dramatically increase mental
workload (Recarte & Nunes, 2003).
This is especially troubling due to the
recent increase in the popularity of mobile
phones (Incisive Interactive Marketing,
2005)
Background Cont.
85% of all mobile phone owners talk on
their phones at least occasionally while
driving (NHTSA, 1997)
21% of crashes or near crashes reported
by respondents involved at least one
driver using a mobile phone (Seo & Torabi,
2004).
Previous Work
Strayer & Johnston (2001) found
participants who used a mobile phone
(both hand-held and hands free)
performed worse in a driving task
compared with participants who passively
listened to radio broadcasts or books on
tape.
Thus the ‘hands’ aspect is not what
degrades driving performance
Previous Work cont.
Strayer et al. found that people that talking
on mobile phones in a driving task were
more likely to experience ‘looked-butfailed-to-see’ errors (2003)
Crundall et al. (2004) found that people
talking on mobile phones have shorter
fixation durations – which may account for
‘looked-but-failed-to-see’ errors
Previous Work cont.
It has been well established that talking
while driving degrades driving
performance.
It is not known, however, which aspects of
‘good’ driving are affected when talking on
a mobile phone while driving (Gugerty,
2004)
Purpose
 Engaging in TMWD increases driving errors as
well as ‘looked-but-failed-to-see’ errors, it is not
known how visual search strategies are modified
according to the specific driving task.
 The current study sought to quantify if/how
visual search patterns change while engaging in
a mobile phone conversation as well as
combined with potentially hazardous driving
situations
Participants
16 (11 female) Clemson University
undergraduate students
20/20 or corrected to 20/20 vision
A valid drivers’ license
At least 2 years driving experience (M =
3.5 years).
One person was not able to participate
due to poor tracking
Apparatus
Tobii 1750 eye tracker
Sampling rate of 50 hertz
1280 x 1024 display 17 LCD screen
Design
Between subjects, 2 x 2 design.
8 people (3 male, 5 female) participated in
the mobile phone condition
8 people (2 male, x 6 female) participated
in the non-mobile phone condition.
All participants viewed 12 trials with 4
vehicles and 12 trials with 7 vehicles in the
driving scene.
Development of the Driving Simulator
C++, OpenGL, SDL
Dynamic ROI generation
Synchronization of frame rate and eye
tracker
Mirrors
Language task
Pimsleur Japanese language learning
compact disk set for beginners
3 language aspects:
Listening
Repeat
Generate
Synced to begin and end with each driving
scene
Procedure
Participants were given instructions
Practice trials
Calibration
View trials (people in the mobile phone
task ‘spoke’ simultaneously)
Answered a question about what occurred
during the scene
Confidence in their response
Procedure
After the completion of the 24 trials,
people responded to a questionnaire
about their attitudes and thoughts about
mobile phones
Results
People on the mobile phone answered
fewer questions about the scene correctly
F (1, 14) = 49.594, p < .001 (37% vs. 68%)
People in the non-mobile phone group
were more confident in their responses F
(1, 200) = 23.314, p < .001. (4.03 vs. 3.18)
Overall people answered more questions
with 4 vehicles correctly than with 7 F (1,
380) = 11.861, p = .001. (60% vs. 44%)
Results
Survey Results
 All participants owned a mobile phone
 On average, participants reported using their
mobile phone ‘sometimes – often’ while driving
 4 participants reported using their phone nearly
every time they drove.
 All felt others’ driving performance is degraded
while TMWD
 However, 7 of the 16 participants felt their driving
performance was only degraded slightly or not at
all
Eye Data Analysis
Removed bad data
Velocity filter to determine fixations and
durations
ROI output from driving simulator
compared with fixations
Eye Data Results - Overall
 Mobile Phone
Fewer total valid points
Percentage of fixations
of total eye points were
not different
 No Mobile Phone
Larger number of total
fixations
The spread of the
fixations were not
different
Eye Data Results – ROIs over whole task
 Mobile Phone
Less time spent in the
ROI
Duration of fixations
was less (supports
looked-but-failed-to see
hypothesis)
 No Mobile Phone
More fixations in the
ROI
Eye Data Results – ROIs during the event
 Mobile Phone
Less time spent in the
ROI
Duration of fixations
was less (supports
looked-but-failed-to see
hypothesis)
 No Mobile Phone
More fixations in the
ROI
Discussion
Language task:
Controlled speed of conversation
Interest level
Etc.
Low-fidelity vs. high fidelity simulator
Eye-data ‘thinking’ phenomenon
Conclusions / future work
People may not be aware of decreased
performance when TMWD
Repeat the expt. with a more ‘involved’
task
Examine the validity
of the language task
Questions!