What's this about?
Data Colada, the forensic-statistics blog that has repeatedly caught fabricated data in behavioural science, presents evidence of fraud in an influential study about procrastination.
How do you detect fabricated data from a published paper?
By looking at properties real data has and invented data usually doesn't. Genuine measurements have characteristic digit distributions, plausible variance, and clustering that follows from how the study was run. Fabricated numbers tend to be too uniform, too smooth, or contain duplicated blocks from copy-paste. None of it requires access to the lab — the anomalies are visible in the published tables, which is why this kind of forensic work has been so productive.
Why is it worth noticing?
HN 74 points, 65 comments
- Data Colada's track record is why this lands: they have been right before, at cost to themselves.
- Behavioural science findings propagate into management advice and popular books long before anyone audits them.
- It arrives the same day as an item about AI-generated papers appearing under real academics' names — different mechanisms, same erosion.