Graham and Patrick led a technology coffee morning focused on their use of the Tobii eye tracker. This coffee morning differed from Jonathan and James’ presentation on eye tracking methods and methodologies in that it was more hands on with a focus on future possible uses of the equipment.
Patrick began by describing how they had used the Tobii eye tracker for analysing websites, finding out what worked and what did not work by seeing how people are actually looking at the site. They very quickly moved onto a demonstration of the eye tracker in practice. The Tobii is a great improvement on previous models. I took part in one of the earlier studies, but the eye tracker was so sensitive that we had to sit in a darkened room which was hardly conducive to feeling relaxed and using the system in a normal manner. The Tobii can be used in bright ambient light and is quite small and neat, looking just like any other flat screen.
Keren offered to be the guinea pig and sat in front of the eye tracker. The screen she was looking at was reproduced on a large display behind her, with the movements of her eyes (the tracks) reproduced real time. They first needed to calibrate the device and, as with most technology, the system failed and eventually they had to switch it off and reboot. Never work with children, animals or technology 🙂
The eye tracker can measure where people look first, how long their eyes rest on a particular point and where the eyes scan around. Highlights the connections between the different parts of the site that people are looking at.
Software produces quite easy to interpret statistics on this. However the interpretation is up to you and depends on your research questions.
Although a Anne Adams has used the Tobii eye tracker to research how people use Second Life, which is a different type of interface, Patrick and Graham have focused on websites. They pointed out that when using eye tracker for website testing purposes. It’s best to have a fairly stable site. For example, if you have a twitter feed or something that changes regularly, the eye tracker will not combine different people’s views of that page. You can combine them manually if you want.
We watched Keren exploring an OU site and then replayed the track afterwards. It was very interesting, with people asking Keren what she had felt at particular points in the recording. For example, it was clear that she had read the terms and conditions, she explained why, but the rest of the text she had merely scanned through. It was possible to ask her why she’d looked at a particular spot and repeatedly returned to it, and why she kept returning to the home page. This highlighted a flaw in the navigational menu in that she felt it was taking her away from where she wanted to be in the website and so had to keep returning.
It was clear from this experience that you can gather a lot of data in a very short time, and that the data consists of both the eye tracking statistics themselves but most importantly, supplemented by a post-interview or discussion with the participant. As Jonathan had pointed out, talk-aloud strategy also helps.
Graham then showed how the data is represented. Graham displayed several people’s tracks on the same image, but pointed out that it’s best to use a website that is in a fairly stable site. For example, if you have a twitter feed or something that changes regularly, the eye tracker will not combine different people’s views of that page. You can combine them manually if you want, which is not difficult to do. However in this version it was noticeable that all the eye tracks were offset to the right. Possibly due to the manual combining of the tracks that Graham had had to do.
Graham demonstrated how eye tracker will overlay several peoples recordings of looking at a single page. You can select a particular time segment to focus down on something of interest.
You have several options – web option, probably also one that caters for second life (video). With web option, if you have a small video on the web page that ends up showing as a blank if people choose to play it. This doesn’t look all that good in the playback.
You can also create an area of interest around a particular part of the screen and extract data related to that specific area.. You can either set these things up before you start or you can create them in response to the data you have collected.
- Count (how many times your eye lands on a partiular place)
- Gaze plot
- Order in which you look at diff parts of the screen
One finding that emerged very quickly in Patrick and Graham’s experiments was that participants found one feature of the website very frustrating. Two participants found that when they clicked on certain links that opened a new window which they tried to resize but couldn’t do it. Kept returning to it. That simple thing caused frustration. Quick feedback to website designers.
You can export to SPSS and do quite a lot more than Patrick and Graham have had time to describe during this coffee morning.
In contrast to Jonathan’s presentation in which he emphasised how quickly you can generate large amounts of data, Patrick and Jonathan focused in on how you can generate useful small amounts of data during very short sessions with the eye tracker. They discussed using it at a forthcoming Learn-About Fair, getting people to sit for a coupl of minutes to track how they use a particular website to get some fast feedback. Other people in the audience described their use of Tobii for:
- Using Tobii to see how different visualisations of data sets work – if one visualisation is better than another for communicating the information.
- Lab testing on cloudworks site. Taking it to workshops and using with academics with academics who are not that experienced with web 2.0. So would like to experiment with people with different levels of experience.
- Testing whether peoples poster displays worked. Using posters on the eye tracker.
Overall, I was quite surprised at how agile a tool the eye tracker seemed to be. You are not obliged to collect vast amounts of data. Indeed, it seems very well suited to quite short data collection stints – delivering usable and useful data. I liked the innovative idea of putting up posters on the eye tracker to see how well they performed. You could use this technique for comparing things that are presented using different media. An hour well spent I feel.