Peer-reviewed Webcam Eye Tracking for Research

Unlocking new and innovative eye tracking research methods is essential for staying at the forefront of any industry or field. Webcam eye tracking is an important tool that researchers can leverage for just that. By utilizing this cutting-edge technology, researchers can implement innovative research and gain more understanding of the processes behind attention, decision-making and behavior. However, it is important to utilize webcam-based eye tracking that has been tested and peer-reviewed.

With the ability to track eye movements, gaze patterns, and fixations in a naturalistic setting, Labvanced’s webcam eye tracking offers a viable approach to quantifying participants’ attention. Moreover, Labvanced’s webcam-based eye tracking technology can be easily implemented remotely, making it a cost-effective solution for researchers all around the world who are interested in using this method in their upcoming projects, all the while protecting participant privacy while being GDPR compliant.

Research paper published by the Labvanced team comparing the webcam eye tracking technology’s accuracy with an industry standard, EyeLink 1000.

Validity and Accuracy of Labvanced’s Online Eye Tracking

• Compared to the EyeLink hardware-based system, Labvanced’s webcam eye tracking has an overall accuracy of 1.4° and a precision of 1.1° and an error of about 0.5° larger than the EyeLink system.
• When visual targets are presented in the center of the screen, both accuracy and precision improve to 1.3° and 0.9°, respectively. This is an important finding given that numerous psychology experiments tend to display stimuli in the center of the screen.
• Findings also showed that accuracy remained consistent across time. For free viewing and smooth pursuit tasks, the correlation was around 80% between Labvanced and EyeLink gaze data.
• For a comprehensive summary of all metrics included in the peer-reviewed paper and detailed information about each eye-tracking system with different tasks, please refer to Table 5 in the paper.

The figure below shows one participant's raw gaze data, showing the gaze data's alignment during the Smooth Pursuit task, separably for the X- and Y- axes.

Fig 1. Graphs from the research paper (corresponds to Fig.7 in the publication), a visual demonstration of how this correlation looks like, in a Smooth Pursuit Task, showing the overlap between the data points between Labvanced (blue dots) and EyeLink (red dots). For free viewing and smooth pursuit tasks, the correlation was around 80% between Labvanced and EyeLink gaze data.

Below are heatmaps with two scatter plots from the participants for two different trials. The left-red represents EyeLink and right-blue Labvanced’s webcam-based eye tracker. The corresponding stimuli image is used as the backdrop for the heatmap in each task.

Fig 2. Heatmaps from the research paper (Fig.8 in the publication).

The values from 0 to 1 in the color bar represent the normalized density values of the kernel density estimate. From the given graphics, we could conclude that the gaze data are assigned roughly to the same objects in the image; however, there is a higher variance across the webcam gaze data distribution.

Effect of Artificial Intelligence Helpfulness and Uncertainty on Cognitive Interactions with Pharmacists: Randomized Controlled Trial

In this study, researchers utilized Labvanced’s webcam eye tracking to determine how pharmacists visually process AI-related information in a medicine verification task. An AI prototype was presented in the context medicine verification where the AI would provide different analyses of the likelihood that the two medications (the pill and the pills in the bottle) were a match. Participants were asked to determine if medication fill images matched reference images, both with and without AI assistance. The AI provided either "black box" help (match status indicators) or "uncertainty-aware" help (match status indicators plus a confidence histogram).

Results:

• When AI-generated regions were present, pharmacists shifted 19-26% of their fixations to them, indicating they considered AI advice in their decisions.
• AI assistance did not decrease fixations on fill images, which remained the main area of focus.
• Unhelpful AI advice was associated with pharmacists spending more time looking at reference and fill images, suggesting increased cognitive effort.
• Showing AI uncertainty resulted in longer cognitive processing times, as seen in increased dwell times on original images.

Advantages of Webcam-based Eye Tracking

Incorporating Labvanced’s webcam eye tracker in your next research project comes with the following benefits:

• Non-invasive: Uses existing devices, providing a more natural and unobtrusive experience.
• Cost-effective: Generally more affordable and accessible than traditional setups.
• Easy to use/setup: Activated with a few clicks in Labvanced, no coding required.
• Customizable calibration: Offers extensive options to tailor calibration to specific research needs.
• GDPR compliant: Processes image data locally, transmitting only gaze-related data such as numeric gaze coordinates.
• Flexible and scalable: Easily deployed remotely for studies with geographically diverse participants.
• Accuracy: A peer-reviewed technology, validated with testing.