Move to Improve: Promoting Physical Navigation to Increase User Performance with Large Displays

Move to Improve: Promoting Physical Navigation to Increase User Performance with Large Displays

Robert BallVirginia TechNovember 10, 2006

Motivation: Cuban Missile Crisis: September/October 1962

U-2 Spy plane takes high-resolutionpictures of parts of Cubahttp://www.gwu.edu/~nsarchiv/nsa/cuba_mis_cri/1.jpg

Picture taken of SA-2 missile construction sitehttp://www.gwu.edu/~nsarchiv/nsa/cuba_mis_cri/1.jpg

Analysis of pictures

U-2 Spy place takes high-resolution pictures. Analyst must

interpret pictureswith this:

Why: physical navigation

Inside the CIA’s National PhotographicInterpretation Center (NPIC), Washington D.C., 1962. http://www.gwu.edu/~nsarchiv/nsa/cuba_mis_cri/7.jpg

Related Work

Games!

Related work – other domains and applications

Physical navigation related workHMDHead-mounted display

CAVE

Experiment Motivation

Research Questions: Do users prefer physical navigation with

large, high-resolution displays? Why? If so, does this result in improved user

performance? Is physical navigation truly more beneficial

than virtual navigation in terms of performance time?

Experimental Design – data and visualization 3,500 houses for sale in Houston, TX Semantic zooming scheme, zooming only resulted in

more information being displayed.

To see all of the houses with all the details shown would require about a 100-monitor display.

Experimental Design – display, interaction, and tracking 24 tiled monitors

Wireless mouse Vicon system

Experimental Design – tasks

4 tasks: Navigation Search Pattern finding Insight finding

Insight task: Used paper, pen, and stand. No correct answers.

Other tasks: Walked and spoke answer. Only one correct answer.

Within-subject design

Between-subject design

Experiment results - performance

2-way ANOVA: main effects fordisplay width (F(1,1324)=20.56, p<0.01) task type (F(2,1324)=77.05, p<0.01)

Tukey HSD analysis showed different task types were all in different groups, so:

TaskMain effect of display width

navigation (F(1,508) = 118.9, p<0.01)

search (F(1, 762) = 38.18, p<0.01)

pattern finding (F(1, 90) = 3.53, p=0.06)

Experiment results – performance, cont.

Performance Times

0

5

10

15

20

25

30

1 2 3 4 5 6 7 8

Display width (in number of columns)

Pe

rfo

rma

nce

tim

e (

s)

Navigation Search

TaskMain effect of display width

navigation (F(1,508) = 118.9, p<0.01)

search (F(1, 762) = 38.18, p<0.01)

pattern finding (F(1, 90) = 3.53, p=0.06)

Experiment results - Performance

We found no significant results based on the level of insight for the fourth task.

So, we focus on results for the first three tasks in this section.

Virtual Navigation Analysis – zooms 2-way ANOVA: main effects for

task type (F(3,1400)=416.2, p<0.01)display width (F(1,1400)=34.8, p<0.01)near-significant interaction of task type and

display width (F(3,1400)=2.4, p=0.06).

Virtual Navigation Analysis – zooms per task

Tasks - MetricsMain effect of display width

navigation - zooms (F(1,508)= 144.6, p<0.01)

navigation - panning not significant

search - zooms (F(1,762) =114.1, p<0.01)

search - panning (F(1,762) = 26.7, p<0.01)

pattern - zooms not significant

pattern - panning (F(1,90) = 7.8, p<0.01)

Number of zooms

0

5

10

15

20

25

1 2 3 4 5 6 7 8

Display width (in number of columns)

Nu

mb

er

of z

oo

ms

Navigation Search Pattern

Virtual Navigation Analysis – pans

2-way ANOVA: main effects for task type (F(3,1400)=301.3, p<0.01)display width (F(1,1400)=63.86, p<0.01) interaction of task type and display width

(F(3,1400)=17.22, p<0.01).

Virtual Navigation Analysis – pans per task

Tasks - MetricsMain effect of display width

navigation - zooms (F(1,508)= 144.6, p<0.01)

navigation - panning not significant

search - zooms (F(1,762) =114.1, p<0.01)

search - panning (F(1,762) = 26.7, p<0.01)

pattern - zooms not significant

pattern - panning (F(1,90) = 7.8, p<0.01)

Number of pans - navigation and search tasks

0

50

100

150

200

1 2 3 4 5 6 7 8

Display width (in number of columns)

Nu

mb

er

of p

an

s

Navigation Search

Number of pans - pattern task

0

200

400

600

800

1000

1200

1 2 3 4 5 6 7 8

Display width (in number of columns)

Nu

mb

er

of p

an

s

Physical Navigation Analysis

Douglas-Peucker algorithm used to guarantee that what we were analyzing was actual movement - not jitter.

Physical Navigation Analysis, cont.

2-way ANOVA for total X distance: main effects for task type (F(3,1400)=75.1, p<0.01)display width (F(1,1400)=24.1, p<0.01) interaction of task type and display width

(F(3,1400)=4.0, p<0.01)

Physical Navigation Analysis – per task

Task Main effect of display width

navigation not significant

search (F(1,762) = 4.52, p=0.03)

pattern finding (F(1,84) = 16.62, p<0.01)

Total Distance in the X Direction - search task

020406080

100120

1 2 3 4 5 6 7 8

Display width(in number of columns)

Mo

vem

en

t in

inch

es

.

Total Distance in the X Direction - pattern task

0

100

200300

400

500

1 2 3 4 5 6 7 8

Display width(in number of columns)

Mo

vem

en

t in

inch

es

.

Visualization of physical movement

Experiment conclusions

Do users prefer physical navigation with large, high-resolution displays?

YES

When faced with a choice of physical or virtual navigation, 100% of the participants

chose only to physically navigate.

Experiment conclusions

If so, does this result in improved user performance?

YESPerformance Times

0

5

10

15

20

25

30

1 2 3 4 5 6 7 8

Display width (in number of columns)

Pe

rfo

rma

nce

tim

e (

s)

Navigation Search

Experiment conclusions

Is physical navigation truly more beneficial than virtual navigation in terms of performance time?

YES

Virtual navigation has a greater negative correlation on performance than physical navigation. The number of zooms correlated with performance with a

correlation coefficient of 0.69. The number of pans correlated with performance with a

correlation coefficient of 0.68. However, physical distance traveled did not significantly

correlate with performance.

Future work

Do the results from this study extrapolate to 3D data? To non-geospatial data?

How does fatigue/longitudinal use of such displays affect the results of this experiment?