Use a description task for dense BubbleView studies

Add a description task when the visual has enough semantic or textual content to describe and you need reliable importance maps from few participants. For example, ask viewers to click and type a short description for an information visualization or webpage instead of only free-viewing it, and use the submitted text to screen out poor trials.

A description task changes where viewers spend effort. They click more selectively on regions that help them complete the task, so informative regions emerge with fewer participants.

Mechanism: Typing while clicking adds an effort barrier. That barrier reduces casual exploration and concentrates clicks on task-relevant areas that support the written description.

Evidence: BubbleView performed best on information visualizations with a description task, and on webpages the description task produced better fixation approximation than free viewing for small participant counts while also enabling description-based quality control (Kim et al., 2017).

Notes: The paper recommends BubbleView especially for defined tasks rather than unconstrained viewing.

• User Goal: Approximate eye fixations or recover the most important regions of a static visual.
• Task: Run a remote BubbleView study with limited participants.
• Data: Dense static images with readable text, labels, or a main message that can be summarized.
• Chart Setting: Blurred image with click-to-reveal bubbles and a text field for the response.
• Success Criterion: Important regions stabilize quickly and poor-quality responses can be filtered.

Break it when: The image cannot be objectively described or requires outside context to interpret. Why: The task stops matching the visual, so description quality becomes hard to judge and the clicks no longer cleanly reflect importance.

Sacrifice: The task takes much longer and costs more than short free-viewing. Risk: The task can bias clicks toward the regions needed for the written summary, especially text-heavy regions. Mitigation: Reserve the description task for visuals with a clear message and enough content to describe.

Mistake: Using free-viewing on dense visuals when you only have a small participant pool. Why it fails: The click maps converge more slowly, so the important regions remain noisier for the same sample size.

Failure Sign: Early click maps stay diffuse and the written responses are vague or incomplete. Quick Check: Read a small batch of descriptions and confirm that they mention the main content of the image. Stronger Test: Compare hotspot stability from the first 10-12 participants under the description task versus a free-viewing pilot.

• Replace a free-viewing instruction with a click-and-describe instruction.
• Require a minimum response length before participants can continue.
• Remove trials with poor descriptions before building the click map.
• Give participants unlimited or longer viewing time so they can alternate between clicking and writing.