[Sigia-l] eye tracking?

[Jared M. Spool]

At 09:07 PM 1/9/2006, Christopher Fahey wrote:
>I'm perfectly happy and eager to hear rebuttals to this skepticism, as I 
>have zero actual experience with the technology and techniques.

Actually, Chris, I have a *ton* of experience with it and I think you're 
right on.

As one of the early folks (maybe even the first) who looked at web usage 
with an eye tracker, we found it a fun toy that could really occupy a lot 
of our time, both in data collection and analysis. Though software has 
improved slightly since we first started looking at it, it hasn't solved 
the big issues.

People think that an eye tracker measures what a user sees. In fact, it 
only measures what their eyes gaze at. There's a significant difference.

The more we played with it, the more we realized that, because a user gazes 
at a design element doesn't mean they see it. We saw many instances where 
the user gazed straight at something for many seconds but not recall ever 
seeing it.

One of the more interesting unexplained phenomenon was that we noticed that 
many users could land the mouse pointer in the scroll bar and successfully 
scroll *without ever gazing at the scroll controls*. Our inference was that 
they were attaining the scroll controls with their mouse using their 
peripheral vision. If this is true, that means that users actually can see 
things *they don't gaze at."

So, if users don't see things they do gaze at and don't gaze at things they 
see, what does an eye tracker actually tell you? This is why you haven't 
seen any research from us recently on this topic.

The eye tracker lets you *observe* where the gaze of the user's eye goes 
relative to the screen. However, it does not help with inferences. (See 
http://tinyurl.com/9kxou for why this distinction is important.)

One of my favorite examples of poor inferences from eye tracker data is 
found in the Eye tools blog. (Eye tools makes software and sells a service 
based on that software, so they are very pro-eye tracker.)

In http://tinyurl.com/7vabo , they claim that they "improved" the SFPD web 
site by using eye tracker data to guide their decision. Look real close at 
the before and after. Their inference was that people weren't looking at 
the navigation, so what did they do to the site? They put a mongo-big 
useless image in the portion of the page users *were* looking at, to force 
them to look at the now only useful part of the page. From their site:

>The moral of the story: A change on one part of the page can impact other, 
>unrelated elements on the page. The right navigation bar was used 
>completely differently on the new re-designed website because the content 
>to the left of it changed.
>
>Is this good or bad? Ultimately, that's up to the client to decide as it 
>relates to the business goals of the page. However, in this case, it was 
>good ­ it enabled visitors to quickly locate the specific content they wanted.

Huh. They didn't measure if visitors actually locate the content they want 
more quickly. The only measured if people's brief eye gazes went to what 
they believed were the proper portion of the screen. That's quite a jump...

Eye tracking can be useful for some types of measurements, but as Chris and 
Stewart have mentioned, it's not particularly a useful tool for day-to-day 
screen design.

Jared


Jared M. Spool, Founding Principal, User Interface Engineering
4 Lookout Lane, Unit 4d, Middleton, MA 01949
978 777-9123   jspool at uie.com  http://www.uie.com
Blog: http://www.uie.com/brainsparks