Eye Tracking: A Method for Online Research
Remote eye tracking is a powerful method for conducting online research on websites and is quickly rising in popularity as a method. Discover our webcam-based technology, how it works, as well as the kind of data reported, including website eye tracking heatmaps.
Contents:
Background
Remote eye tracking is a quickly rising technology as a promising (and easy) method for measuring attention online just by using a webcam. Our webcam-based eye tracking framework (which has, up to this point, been available for studies conducted solely within Labvanced for psychology-based research experiments) has been tried and tested for its accuracy and usability.
We have been active players in this field for the past several years, largely driving the innovation and accuracy of this technology. Now, we are taking the next step by expanding to make it possible to apply webcam-based eye tracking to any website - via iFrame technology - so it can be used as a method for website research in various use cases, including UX. Additionally, by combining eye tracking on websites with mouse tracking, we aim to provide useful data and intuitive analytics that capture attention and hand-eye coordination patterns, all under the span of the ELISE & EU Horizon 2020 research and innovation programme.
Applications
By quantifying attention with eye tracking online and combining behavioral metrics (such as clicking), organizational decisions (like marketing and UX design) can be confidently made by prioritizing what site elements are most interesting for the target audience. This will help you minimize the search time that visitors spend on your website while maximizing the time that visitors spend reading text or interacting with content.
- Free-browsing: Users are prompted to freely browse a website without any specific instructions. During the free browse, their gaze is quantified with remote eye tracking in order to determine which elements are attention-grabbing on the site.
- Find & click: Users are asked to find a certain website element and click on it. Gaze is measured throughout this process to determine what parts of the website are paid attention to while looking for something specific.
- A/B testing: By showing two different site versions with changes to visual elements and design, you can determine what version of the A/B test draws most attention and how this is related to behavioral inputs like clicking certain areas.
- UX/UI design and research: UX/UI researchers and designers can now utilize webcam-based eye tracking as a part of their quantitative data collection to objectively study user flows, friction points, and the overall usability of their designs.
- General website optimization: Eye tracking used to be a technique that was only available for those who could afford expensive hardware. Now, it can be used by any website owner to optimize the performance of their site.
- E-commerce: Website owners with e-commerce shops can utilize eye tracking to see how users are looking at their product pages, as well as what areas capture attention prior to purchasing a product.
- Online behavior research: Researchers that are interested in studying website behavior in general can also utilize PageGazer to perform experiments and quantify attentional processes and decision-making.
Ultimately, PageGazer is a powerful solution that addresses the shortcomings of existing solutions by its novel approach of combining mouse tracking and eye-tracking with online website data all the while reporting accurate data while being an easy method to implement and use.
Reported Data
The data that is reported is easily aggregated across the participants due to our website parsing technology. This will allow you to draw conclusions (over the average of your participants) for particular AOIs. For example, users are likely to look at Area A before clicking on Button B. Such findings are the basis of the insights provided by PageGazer and have implications for decisions you make as a business, prioritizing certain types of content over others.
The type of data collected by employing eye tracking online to assess your website includes following:
- Time to first fixation
- Gaze paths
- Gaze duration
- Scanning behavior
- Gaze heatmaps
- User states (distinguishing between cognitive states like reading, searching, image viewing, browsing, and video watching)
Features
Virtual Chinrest
Eye tracking as a method requires the participant to keep a fixed head position as movements should be limited. Traditionally, eye tracking works by using specialized hardware and participants stay still by physically placing their chins on a stand, ie. a chinrest. For remote eye tracking, a webcam is utilized and a ‘virtual’ chinrest detects when a participant moves too much and, as result, the experiment pauses and on-screen instructions appear prompting the participant to return to an acceptable position. Thus, the virtual chinrest is crucial for ensuring that quality data is recorded when using eye tracking online and as a remote method.
Calibration
Calibration is an important step before an eye tracking study begins. Calibration involves ‘prepping’ the eye tracking neural network to be calibrated to the user in order to collect accurate measurements and gaze coordinates. The calibration process can be 1-minute long, 3-minutes long, or 5 minutes long. The longer the calibration process, the more accurate the data collected.
The 1-minute calibration process consists of focusing on specific dots across various sections of the screen for several seconds each. The 3- and 5-minute calibration options also include this calibration task, but they also include a smooth pursuit task where the participant is asked to follow a moving dot across the screen.
Eye Tracking Settings
In PageGazer, these two factors can be specified in the eye tracking settings in the study editor, as shown below. The standard settings include a 3-minute calibration process and a ‘very loose’ virtual chinrest. This is the default setting because it provides relatively good accuracy levels while not being too demanding of the participants with regards to how much time it takes to complete.
You may want to change these settings based on the nature of your study. For example, you may want to choose the lengthier calibration process (5 minutes long) for a study that takes 30 minutes to complete whereas a short calibration process (1 minute long) might be more optimal for a much shorter study.
These settings ultimately allow you to control important aspects of eye tracking for online research.
Accuracy
The level of accuracy in a study depends on the strictness of the virtual chinrest as well as the length of the calibration process. Ultimately, accuracy is crucial for choosing eye tracking as a go-to method for online research. Check out this peer-reviewed paper that focuses on our webcam-based eye tracking technology’s accuracy in the journal of Behavior Research Methods which is now being used in PageGazer.
These are some key findings from it:
- The webcam-based eye tracking has an overall accuracy of 1.4° and a precision of 1.1° with an error of about 0.5° larger than the EyeLink system.
- Interestingly, both accuracy and precision improve (to 1.3° and 0.9°, respectively) when visual targets are presented in the center of the screen - something that’s important for researchers to take into account given that the center of the screen is where stimuli are presented in a lot of psychology experiments.
- For free viewing and smooth pursuit tasks, the correlation was around 80% between the data acquired from our webcam-based eye tracking and EyeLink 1000.
- Also, the accuracy was consistent across time.
Overall, these results show that our technology is a viable option for studying the physiological signals of attention by using just a webcam.
GDPR Compliance
As working with cameras is a sensitive topic in general, we value privacy and have developed our webcam-based eye tracking to protect privacy levels at its very core.
How is this done? We fully adhere to GDPR by ensuring that no face/image data ever leaves the participants’ devices. When the eye tracking is working, real-time gaze estimation on the participants' devices. Thus, the only information that is sent to the servers is the predicted eye location in x,y,z coordinates (ie. the numerical value of where the gaze was on the screen). Therefore, the optimal way to secure data privacy (for webcam-based approaches) by design is to process the image data on the participant's device directly and not to transmit video data and not to store them on a remote server which is exactly what the technology behind PageGazer does.
About Labvanced
Our team behind Scicovery GmbH has a strong scientific background and we have conducted several online studies ourselves during our own PhDs. However, until Labvanced was launched in 2017, no online platform was yet able to combine all the advantages of online research into one single application. This is essentially what Labvanced is today, an all-in-one solution for web based research. With Labvanced, you can perform precise behavioral (decision making, reaction times, video recordings, and more.) and physiological measurements (eye tracking, head tracking) while sharing and collaborating with your colleagues, all without requiring knowledge of coding. Since its launch in 2017, over 3000 researchers from all over the world and across various disciplines have used Labvanced, from cognitive psychology to linguistics to clinical psychology and more. There are hundreds of studies running on Labvanced in all directions and with various degrees of complexity. Check out the Publications page for a list of peer-reviewed research papers, reviews, and even dissertations using and citing Labvanced.
While our innovative webcam-based eye tracking is available within Labvanced, the next step was to move it away from this framework and enable it to run on websites. This is where the ELISE grant comes in.
EU Horizon 2020 & ELISE
The European Learning and Intelligent Systems Excellence (ELISE) is a European Network of AI Excellence Centres funded by the European Commission under the Horizon 2020 Framework Programme under the Grant Agreement no 951847. ELISE began in September 2020 and will run for 4 years.
PageGazer was selected to take part of the research and innovation programme for developing its webcam-based eye tracking technology to be functional in all browsers for website and online behavior research while building additional neural networks in conjunction with other innovative capabilities, such as website parsing.
For more information on ELISE, please see www.elise-ai.eu
This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 951847.
