Using eye tracking to determine whether person needsmore information.

25 Jan 2008 - 3:40am
6 years ago
1 reply
424 reads
Joakim Isaksson
2008

Hi Peter,

This sounds like a really interesting project. Are you planning to adapt the information being displayed on the screen based on the satisfaction of the user? I think that could be very risky unless you get a really high reliability in your analysis as automatically changing what is displayed can be frustrating if you don't want it to change.

I would probably not use pupil size data in this analysis as it depends on so many different factors like lighting conditions, the brightness of the stimulus and arousal (which in turn can be affected by numerous things). Unless you run this system in a very controlled environment I think you would have a really hard time getting accurate results.

This also touches on the key problem that you face; you say that you don't want to take into account what content is being shown to the user. It will be hard to determine the user's state of mind based on the gaze pattern as images in different layouts produces gaze patterns that differs a lot from reading patterns, even though the images can be just what the user was looking for. Reading detection is probably fairly simple but is only useful if all content is text based. Will you be able to use information on the navigation structure in the algorithm?

Best regards,

Joakim Isaksson

----------------------------------------------------------------
Joakim Isaksson
Support Engineer
E-mail:    joakim.isaksson at tobii.com
Web:       www.tobii.com

Comments

30 Jan 2008 - 7:35am
Peter
2008

Hi Joakim,

thank you for your very helpful answer, especially concerning the
drawbacks of the pupillometrics.
Also the point with the navigation structure. It is yet unclear,
whether I could use this as additional parameter. I would be happy,
however, to be independent from that.
I hoped that there already were some findings about how to classify
a gaze path as "normal viewing" and as "seeking". - Do you know
whether anyone already examined this?

Best regards,

Peter.

Syndicate content Get the feed