🔬 SciVisCap Human Evaluation

About this study

You will evaluate captions for scientific visualization (SciVis) figures from published visualization papers. Scientific visualizations represent spatially grounded scientific data using techniques such as isosurface rendering, direct volume rendering, and flow visualization.

For each item, you will see a SciVis figure, one or more paragraphs from the paper that reference the figure, and three candidate captions. Your task is to rank the three captions from best to worst according to five criteria.

The study includes 30 figures and is expected to take approximately one hour. Your responses are saved each time you click the Next button, so you may close the page and resume later using the same login.

How to judge

Use the SciVis figure and its referencing paragraphs as evidence to rank the captions on each criterion.

Data

We collect your name, institution, and email to track study completion and contact you about compensation. Ratings will be analyzed in aggregate and reported anonymously. Contact [email protected] with any questions.

Figure

📖 Referencing paragraph(s)

Caption A

Caption B

Caption C

Rank the three captions from best to worst for each criterion.

Quality. How clear, grammatical, and appropriately concise is it as a figure caption?
Best
Worst
Fidelity. Does the caption accurately describe what is visually supported by the rendering without asserting content that is not shown?
Best
Worst
Correctness. Is the interpretation of the depicted structures or phenomena scientifically correct?
Best
Worst
Comprehensiveness. How completely does the caption cover the important visual and scientific information needed to understand the figure?
Best
Worst
Takeaway. How effectively does the caption convey the main point, finding, or conclusion that the figure is intended to communicate?
Best
Worst