Title: Improving Reproducibility of Eye-Tracking Studies

Author: Ana Martinovici

Abstract:
Eye-tracking studies are broadly used to obtain overt measurements (i.e.,
eye-movement data) of visual attention. Researchers who use such studies face
multiple layers of complexity due to the use of proprietary software for data
collection, the amount of data, and the lack of guidelines on data processing
and analysis. For example, an eye-tracker records the position of each eye
50-1000 times per second (e.g., typical marketing studies use a sampling rate
of 50-100Hz, while psychology studies use 500-1000Hz). This leads to a large
number of observations per participant, that need to go through complex
processing steps before the researcher can test hypotheses. Most existing
eye-tracking software requires researchers to complete these steps manually,
through a point-and-click interface. For example, researchers use the mouse to
draw areas of interest (AOI) on an image and then use point-and-click menus to
check the number of times that participants looked at that AOI. These
processing steps are almost impossible to reproduce, as most eye-tracking
software doesn’t allow the use of scripts to document the specific options
selected by the researcher. Processing eye-tracking data in a reproducible way
is possible if researchers have the time and experience necessary to build
their own custom-made solutions. This talk provides examples of how R can be
used for eye-tracking data collection and processing (e.g., generating
stimuli, defining AOI, processing of raw eye-samples). These examples are part
of an R package under development that enables researchers to increase the
likelihood of reproducibility for their eye-tracking studies.