Transcript Transformation of Digital Libraries: From Information

Eye Tracking in the Design and Evaluation of
Digital Libraries
Bing Pan, PhD, Geri Gay, PhD, Helene Hembrooke, PhD,
Laura Granka, Matt Feusner
HCI Group
Department of Communication and Information Science
Cornell University
http://www.hci.cornell.edu
Eyes are Windows to the Soul!
Eye Tracking Experiment
ASL’s 504 commercial eye-tracker
(Applied Science Technologies, Bedford, MA)
Eye Tracking Measurements

Micro-level of information processing
– Fixation:
 Indication of attention
 Intensive information processing during fixations;
– Saccade:
 Indication of movement of attention
 Information processing is suppressed.
Eye Tracking Example
Scanpaths
Fixations
Saccades
Previous Research
Eyes are drawn to informative areas (Rayner, 1998);
 Complex information leads to longer fixation
duration; Fixation numbers are indication of
importance of visual display (Fitts et al., 1950);
 Difficult task leads to longer saccade (Takahashi et
al., 2000);
 Different tasks lead to different eye movement
behavior (Hayhoe et al., 2002).

Study I: Eye Tracking Research on
Web Pages
30 Subjects, 22 web pages from 11 most
popular web sites from four categories (news,
shopping, search, and business);
 Half of subjects are asked to memorize the
content;
 30 seconds of viewings;
 Demographic data and recall were measured.

Area of Interest (Lookzone) Analysis
Where are the visually salient areas?
Do people look at those?
Where do people look on a web page?
Illuminating visually salient areas...
Using illumination as another representation...
Study I Research Results
(under review):
– Males spent more time on fixations; females
spent more time on saccades;
– The adjustment of people’s eye movement
behavior on different pages;
– Complexity of web pages leads to complex
scanpaths…
Study II. Eye Tracking in DL
MetaTest Research

How users of DLs use metadata and process
metadata?
– Test on three conditions:
 Records with descriptions;
 Records with Metadata;
 Records with both descriptions and metadata.
DL collections with descriptions…
DL collections with metadata…
DL collections with
both descriptions and metadata…
Study II MetaTest
Initial Results
– Titles and sources are the mostly viewed
metadata;
– The first few sentences in the description are
read more carefully; the rest of them are
skimmed;
– Before selection, a re-visit of the records for
confirmation;
– Subjects focus on descriptions when both
descriptions and metadata are on the same page.
Application in DL Design and
Evaluation





Which areas attract people’s attention?
How can we capture their attention?
Which areas are left out by viewers’ eyes?
How can we design individualized
interface?
How does their attention switch from one
page to another?
What is the average scanpath on a web
page?
Conclusion
Eye tracking can be a powerful tool in DL
evaluations and usability testing;
 Interface can be improved based on eye
tracking results;
 Comparison between alternative interfaces.

Future Directions
From information retrieval to learning
community;
 Will people use social navigation tools by
attending to those social elements?
 Testing the prototypes before actually
spending money designing them.
