Niels Henze University of Oldenburg Enrico Rukzio University of DuisburgEssen Susanne Boll University of Oldenburg 100,000,000 Taps Analysis and Improvement of Touch Performance in the Large
Jul 09, 2015
Niels Henze University of Oldenburg
Enrico Rukzio
University of Duisburg-‐Essen
Susanne Boll University of Oldenburg
100,000,000 Taps
Analysis and Improvement of Touch Performance in the Large
Touch me How can it be Believe me The sun always shines on my screen
but sJll...
we cannot see what we touch fingers are bigger than the elements we touch
Work about mobile
touchscreens? special interacJon techniques
Work about mobile
touchscreens? special interacJon techniques performance for different target sizes and posiJons
Y. S. Park, S. H. Han, J. Park, Y. Cho: Touch key design for target selecJon on a mobile phone. Proc. MobileHCI, 2008.
Work about mobile
touchscreens? special interacJon techniques performance for different target sizes and posiJons taps might be systemaJcally skewed
Y. S. Park, S. H. Han, J. Park, Y. Cho: Touch key design for target selecJon on a mobile phone. Proc. MobileHCI, 2008.
Degrees of freedom
different sizes
Degrees of freedom
different sizes different posiJons
Degrees of freedom
different sizes different posiJons different devices
Degrees of freedom
different sizes different posiJons different devices …
Large amount of
touch data game published on the Android Market
Large amount of
touch data game published on the Android Market we inform the player about the study
Large amount of
touch data game published on the Android Market we inform the player about the study just looks like an ordinary game
Large amount of
touch data game published on the Android Market we inform the player about the study just looks like an ordinary game parJcipants get some introducJon
Large amount of
touch data game published on the Android Market we inform the player about the study just looks like an ordinary game parJcipants get some introducJon
Large amount of
touch data game published on the Android Market we inform the player about the study just looks like an ordinary game parJcipants get some introducJon they tap the targets
Large amount of
touch data game published on the Android Market we inform the player about the study just looks like an ordinary game parJcipants get some introducJon they tap the targets we vary targets’ size and posiJon
We collected
data from 91,731 installaJons
We collected
data from 91,731 installaJons > 50 different devices
The ten most common devices
We collected
data from 91,731 installaJons > 50 different devices 120,626,225 touch events
Using the data
we determined the error rate for different sizes
Using the data
we determined the error rate for different sizes and the error rate for different posiJons
error rate
HTC Wildfire
SystemaJc skew
found that taps are systemaJcally skewed
HTC Wildfire
SystemaJc skew
found that taps are systemaJcally skewed determined offsets for each screen posiJon
Samsung Galaxy S
SystemaJc skew
found that taps are systemaJcally skewed determined offsets for each screen posiJon derived how to shid taps to decrease error rate
devices with 480x800 pixels resoluJon
SystemaJc skew
found that taps are systemaJcally skewed determined offsets for each screen posiJon derived how to shid taps to decrease error rate approximate a smooth shid funcJon
devices with 480x800 pixels resoluJon
Experimental
validaJon implemented shid funcJon for 4 resoluJons
control condiJon experimental condiJon
Experimental
validaJon implemented shid funcJon for 4 resoluJons published an update of the game
control condiJon experimental condiJon
Experimental
validaJon implemented shid funcJon for 4 resoluJons published an update of the game randomly assigned players to a condiJon
control condiJon nc=6062
experimental condiJon ne=6139
Shid funcJon
decreases the error rate
Shid funcJon
decreases the error rate increases the precision
Shid funcJon
decreases the error rate increases the precision for all considered resoluJons
Summary interacJve task
implemented in a game touch data from 91,731
installaJons determined error rates and systemaJc skew
derived shid funcJon validated in a controlled
experiment
LimitaJons know ligle about the players experimental manipulaJon only has a small effect unlikely that the funcJon works for other tasks
Niels Henze niels.henze@uni-‐oldenburg.de
University of Oldenburg
Enrico Rukzio enrico.rukzio@uni-‐due.de
University of Duisburg-‐Essen
Susanne Boll susanne.boll@uni-‐oldenburg.de
University of Oldenburg
100,000,000 Taps
Analysis and Improvement of
Touch Performance in the Large
ContribuJon
“App Store study” to build a model for an interacJve task
public between-‐groups experiment for the task
compensaJon funcJon for touches beyond specific
devices, users, and contexts
Interested in the data? hgp://nhenze.net/?page_id=673