CARE’s Behaviour Tracking feature goes beyond collecting annotations it captures how users interact with the platform. When participants opt in, CARE records things like click patterns, page navigation, scrolling, search activity, and even when users switch to another tab. Right now we save these statistics for the annotator and editor, and we are working on implementing behaviour tracking for other parts of the platform as well. Together, these signals give researchers a rich, behind-the-scenes view of the annotation process that surveys and final outputs alone could never provide.
Privacy is at the heart of this feature. Behaviour tracking is entirely opt-in — users see a clear consent checkbox during registration that says something like “I allow the collection of behaviour statistics for research purposes.” If you prefer different wording, you can set your own custom privacy message. If users don’t consent, nothing is recorded. The data is anonymous and only accessible to study administrators, ensuring participant trust is never compromised.
So how does this help research? In many ways. Researchers can study how annotators approach a task whether they work methodically or jump around, where they spend the most time, and when they seem confused or fatigued. This makes it possible to measure task difficulty, compare different study conditions, detect engagement drops, and improve tool design based on real usage patterns rather than guesswork.
The applications extend well beyond NLP. Since all the collected data can be exported as CSV or JSON, plugging it into your preferred analysis tool is straightforward.
In short, Behaviour Tracking in CARE bridges the gap between what users produce and how they produce it, giving researchers the deeper understanding they need to design better studies and ask better questions.