Paintings, Not Noise—The Role of Presentation Sequence in Labeling
2.88T1.5 sourceInteracting with Computers (Oxford Academic)
Source record
Published by Interacting with Computers (Oxford Academic) (T1.5 source). The original is at https://academic.oup.com/iwc/article/38/3/352/7629774?rss=1.
Pipeline notes
The summary and note below are generated by the signal pipeline — they are Beyond Desk’s reading, not quotations from the source.
SummaryAcademic study testing how presentation sequence (by label, by image, random) in ML data labeling tasks affects 176 crowd workers' perceived variety, autonomy, motivation, and performance under self-determination theory. Counterintuitively, by-label sorting was perceived as more varied than random sequencing.
Why it mattersCounters the assumption that randomization boosts variety in labeling work. Useful design lever for anyone structuring human-in-the-loop dataset pipelines.
Cited by
No citations on record.