Commercial Uses of Eye tracking

1
Commercial Uses of Eye tracking Workshop 5th
September 2005 HCI Edinburgh This workshop was
aimed at sharing and defining best practice,
ideas, and unanswered questions on the scientific
and practical use of eye tracking as applied to
commercial UCD.

• Attendees
• Linden Ball (Lancaster University, UK)
• Peter Brawn (Access Testing, Australia)
• Laura Cowen (IBM, UK)
• Stephen Gulliver and George Ghinea (Brunel
  University, UK)
• John McCarthy, Jens Riegelsberger (University
  College London, UK)
• Kara Pernice Coyne (Nielsen Norman Group, USA)
• Sharon O'Brien (Dublin City University, Ireland)
• Tony Renshaw (Leeds Metropolitan University, UK)
• Charlotte Sennersten (Gotland University, Sweden)
• Ekaterini Tzanidou (Open University, UK)
• Natalie Webb (Amberlight, UK)
• Alan Woolrych (University of Sunderland, UK)
• Daniela Zambarbieri (Università di Pavia, Italy)

2
Discussions

• Metrics and Definitions
• Fixations and scan paths within a known context
  (e.g. search) are accepted metrics for
  interpreting data
• Pupil dilation and blink rate however need more
  research to know the right conditions for their
  use e.g. problems with luminance affecting pupil
  dilation
• Future needs
• Complete disclosure of study parameters e.g.
  definitions of fixations. This will enable
  research comparisons.

Scan path
3
Discussions

• Eye Mind Hypothesis
• Generally accept assumption that where people
  look indicates what they are paying attention to,
  or thinking about.
• However, care needs to be taken in respect of
• The context of eye behaviour e.g. a specific
  search task allows more confidence in inferences
  drawn whereas an open brief to look, means more
  factors are likely to influence behaviour, such
  as meaningfulness, visual (bottom-up) cues and
  motivational level
• Expectations, experience and individual
  differences will also influence behaviour
• The role of peripheral vision and pre-attentive
  processing cannot be directly determined by eye
  tracking and need to be inferred from eye
  movement data
• The hypothesis that users might consciously
  frustrate studies (e.g. changing task during an
  exercise) was considered possible but rare.
• Future work is required to establish the
  boundaries of the eye mind hypothesis for dynamic
  displays with variable frame rates e.g. video

4
Discussions

• Tools and Techniques
• Issue How long does calibration hold good for
  and how does this influence the length of
  studies? The feeling was that more work is needed
  to test assumptions around the accuracy of
  calibration over time.

• Study Design
• Quality of eye tracking studies depends on using
  proven study design principles e.g. good
  sampling
• Ecological validity is important e.g. helping
  users to relax and behave naturally in test
  conditions
• Eye tracking was seen to have an advantage in
  that the facilitator does not necessarily need to
  be in the same room as the participant
• Issue How to predict sample size is still not
  clear due to lack of experimental data and
  universal definitions of eye tracking metrics.
• Future work The group thought that there was
  potential for more studies on individual
  differences in eye movement e.g. age, culture

5
Discussions

• Execution of studies
• The consensus was that participants should be
  informed of their eyes being tracked
• Though users of Tobii believed the equipment was
  fairly robust to natural movements during tests,
  all agreed that it was important to get the user
  in a sustainable position for the duration of the
  calibration and test
• Using re-play of eye movements to elicit feedback
  from participants post test was found to be
  useful but may carry the risk of post-hoc
  justification of actions by participants
• Think aloud protocol during eye tracking was seen
  to provide real-time feedback but altered
  behaviour. Post test replay was preferred by some

6
Discussions

• Analysis
• Care needs to be taken using averages e.g.
  averaging different scan-paths may mean losing
  the ability to detect important differences.
• Issues
• Analysis for dynamic data can be difficult e.g.
  Areas of Interest changing frame by frame for
  video
• Achieving synchronisation of eyetracker with
  multi-media stimulus e.g. games
• Paper journals have limited effectiveness in
  showing analysis in more than 2 dimensions e.g.
  changes over time. e-journals could provide a way
  to overcome this.

7
Discussions

• Validity of eyetracking
• Eyetracking is an attractive methodology for
  clients due to its novel and visual nature.
  Eyetracking could also be introduced more to
  clients as an additional analysis.
• Eyetracking is an excellent methodology for
  answering certain questions e.g search
  strategies it is real time and not dependent on
  self-report. It can also offer complementary
  evidence for much other work.
• Issues Eyetracking is expensive due to equipment
  and additional analysis time in testing.
  Consequently, value behind additional expenditure
  needs to be thought through and explained clearly
  to the customer.