Using eye tracking to determine whether person needs more information.

22 Jan 2008 - 7:57am
8 years ago
3 replies
1131 reads

Hi out there,

I am not sure whether this is the right forum, so sorry in advance if
not. But I saw some questions regarding eye tracking so I hope you
could help me. - BTW do you know forums that are completely dedicated
to eye tracking?

Okay, here is my challenge:

I am working since two years with eye tracking in the context of
dialogue systems. Currently I try to build a system which displays
information for several objects (e.g. buildings, touristic info etc.)
and tries to recognize (via eye tracking) whether the user has enough
information displayed on the screen or whether he/she is seeking for
additional information somehow.
So my question is: Is there a way to determine this by simply ana-
lysing the properties of the scanpath (measuring angles between
saccades, transition densitiy for directness of search, or even better
because simpler solely fixation length or saccadic amplitude) without
the underlying stimulus (making AOIs and such)?
The user wears a head mounted eye tracker (SMI iViewX) and watches the
information on a large screen. Interaction is usual mouse interaction,
clicking on objects for getting context information about them.

I know there's much about reading research, so ones maybe could
distinguish whether the person is reading something. But I think also
need to distinguish whether the person is somewhat relaxed because
he/she is content and has enough information or whether he/she is
seeking for something. Maybe pupillometric data could help here
additionally, but I would be happy not having to base the system on
pupillometric data because I have no experience with that and don't
know how robust that would be.

If anyone of you had some ideas how to do this or some hints where to
further read (papers, names of persons who do research with that) I
would be very thankful.

Thank you in advance and hopefully see you some day.

Best regards,



23 Jan 2008 - 10:30am
Harry Brignull

Hi Peter,

It sounds like a very interesting project, but I suspect that you have a very tricky problem on your hands.

You say you are aiming to build a system that %u201Ctries to recognize (via eye tracking) whether the user has enough information displayed on the screen or whether he/she is seeking for additional information somehow.%u201D

When I%u2019ve done qualitative eye tracking studies, I%u2019ve noticed that when someone spends time reading something, it%u2019s almost impossible to deduce what they are thinking. So for example, if a paragraph is very badly written, they might stop and dwell on that paragraph. Afterwards when you interview them about it, they might say something like %u201Cwell, I was wondering what the hell it was getting at... I%u2019m actually quite frustrated%u201D

Equally, if a different paragraph contains exactly the information they are looking for, and is perfectly written, they might stop and dwell on that paragraph. Afterwards they might say %u201CThat was perfect, I found just what I needed, written clearly. I am satisfied%u201D.

This %u201Cgulf%u201D between looking and thinking is the reason why eye tracking is usually paired with retrospective think-aloud techniques. The user themselves is quite a good judge of whether they want more information on a particular topic.

You might find it useful to do some sketching of your %u201Cfantasy%u201D system (ignoring technical issues for now), and run some wizard-of-oz tests. I%u2019m not sure if I understand your proposition fully, but regardless, this approach will be a cheap way of developing a %u201Cbigger picture%u201D understanding of the problem you are trying to address. (i.e. not %u201Cis it technically possible?%u201D but %u201CDo people need a product like this? And what should it feel like to use?%u201D)

Oh, and by the way, there was a good thread on eye tracking on this forum back in November:

Good luck


31 Jan 2008 - 12:00am
Oleg Krupnov

Sorry for a bit of off-topic, but I have always been curious how such systems
handle the fact that different people have different patterns of eye
movement. Namely, it is known that eyes with imperfect sight (such as myopic
eyes) do not perform short and sharp-cut movements freely and easily like
normal eyes, but instead tend to stare at a single point, trying to see a
big area around that point at once. Is this effect noticeable with your

View this message in context:
Sent from the - discussion list mailing list archive at

31 Jan 2008 - 10:18am
Joakim Isaksson

Hi Oleg and Peter,

Oleg, an eye tracker can only determine the physical orientation of the eye which means that for someone with a condition where they use their peripheral vision and don't move their eyes much, eye tracking would be fairly useless.

Peter, I am not familiar with any work that has been done in that field. I would think that very complex research and large datasets would be needed to find any valid conclusions.

Best regards,

Joakim Isakson

Joakim Isaksson

Syndicate content Get the feed