Data Colada
datacolada.org · Group blog · on HN since 2013
Data Colada is a group of behavioral scientists who examine research data for errors and fraud.
What HN saysCommenters point to its posts exposing data fraud in behavioral science, including the Gino case, and praise how it traced manipulation in Excel files.
80HN people shared it
47linked it in comments
44stories submitted
496story points
3 months agolast shared
2013people per year2026
Most-shared posts
- 1Data Falsificada (Part 1): “Clusterfake” – Data Colada20 people · 5 points · first shared 2023 · 1 comments on HN
- 2Evidence of fraud in an influential field experiment about dishonesty18 people · 168 points · first shared 2021 · 51 comments on HN
- 3Data Falsificada (Part 2): “My Class Year Is Harvard”8 people · 4 points · first shared 2023 · discussion on HN
- 4Data Falsificada (Part 4): “Forgetting the Words”5 people · 1 points · first shared 2023 · discussion on HN
- 5Homepage5 people · first shared 2023
- 6Data Litigada: Thank You4 people · 3 points · first shared 2023 · 1 comments on HN
- 71114 people · first shared 2023
- 8MRAN is Dead, long live GRAN3 people · 4 points · first shared 2023 · discussion on HN
- 9Greg vs. Jamal: Why Didn't Bertrand and Mullainathan (2004) Replicate?3 people · 2 points · first shared 2017 · discussion on HN
- 10Groundhog 2.0: Further addressing the threat R poses to reproducible research2 people · 10 points · first shared 2022 · 1 comments on HN
- 11The Impersonator: The Fake Data Were Coming from Inside the Lab2 people · 9 points · first shared 2024 · 1 comments on HN
- 12Exhibits 3, 4, and 5 – datacolada2 people · 1 points · first shared 2023 · discussion on HN
- 13Harvard's Gino Report Reveals How a Dataset Was Altered2 people · 1 points · first shared 2024 · discussion on HN
- 14Strack et al 1988 cartoons.pdf2 people · first shared 2016
- 15DanBlogComment Aug 16 2021 final.pdf2 people · first shared 2021
What HN people wrote when they linked it
pclmulqdq · 2024-10-21
Reading many scientific studies that are done poorly, I am not convinced that you need a lot of formal training in a field to identify bad science. Many frauds and errors are blatantly visible in their data, which should have a given distribution but doesn't, or in their flawed analyses. Many others have several clearly uncontrolled variables (hint: if it's not in the paper, they didn't control it).
People of a given field will also often circle the wagons around bad actors to protect their own jobs and reputations. This introduces a very strong bias in the people who should be the most highly-informed on their own subject, and sadly makes them less trustworthy than people like Data Colada (https://datacolada.org/) who go beyond their own fields.
smcin · 2024-03-15
For excellent ongoing coverage of Francesca Gino and other misconduct cases in academia, esp. behavioral science (such as Dan Ariely), see Pete Judo. Also covers replication failures, data hacking, and much more.
[YouTube: https://www.youtube.com/@PeteJudo1 , https://www.petejudo.com/]
Also: DataColada blog [https://datacolada.org/], who made public the misconduct reported by Gino's graduate student.
prof-dr-ir · 2025-03-09
The New Yorker article from 2023 about this saga is worth a read:
https://www.newyorker.com/magazine/2023/10/09/they-studied-dishonesty-…
Technical details of the fraud can be found in four famous-ish blog posts of the Data Colada group, also from 2023, beginning with:
https://datacolada.org/109
timsh · 2025-02-03
this is so precise. I guess we’ll need a global version of https://datacolada.org/ quite soon to not get hit by a bus in every scientific field
giraffe_lady · 2024-11-19
> https://datacolada.org/109
The original article on this is a really fun read if you've never seen it. They describe how to bust open the excel file and look at the component xml to figure out exactly how the data was manipulated. It's nothing ridiculously sophisticated just a great example of very motivated digging.
andrelaszlo · 2024-09-13
This analysis is simple but brilliant. Some rows appeared out of order. They changed them back to what they should be according to their position, then checked the effects on the distributions:
https://datacolada.org/111
ColoursofOSINT · 2023-10-01
My favorite thing is that Gino is now suing DataColada [1], and basically is claiming that 'they were right in every other case and I supported them, but with me they are lying'.
1: https://datacolada.org/113
dblack12705 · 2023-09-30
In case anyone hasn’t read it, the Datacolada article demonstrating Ariely committed fraud is a great read and extremely convincing.
https://datacolada.org/98
wigl · 2023-09-27
Love it but note that Dan Ariely is a fraud: https://datacolada.org/98
michaelhoney · 2023-08-02
The Very Bad Wizards podcast talked a little about this case in a recent episode (https://verybadwizards.com/episode/episode-263-free-yoel - main topic is also interesting, about a professor not getting hired for what seem like not-great reasons). The podcast notes include a link to the Data Colada post about the Gino fraud: http://datacolada.org/109
rahimnathwani · 2023-07-28
The two linked source posts are excellent:
http://datacolada.org/109
http://datacolada.org/98
This latest one has great sleuthing examples:
http://datacolada.org/112
Their analysis suggested that the data (originally collected using Qualtrics) had been modified, i.e. that the data in the data files would not match the original data (still stored on Qualtrics' servers).
It turned out they were right:
We have received confirmation, from outside of Harvard, that Harvard's investigators did look at the original Qualtrics data file and that the data had been modified.
cepth · 2018-07-22
Again, academia/research is not my profession. But, some cool efforts in this area include osf.io, which is trying to be the Arxiv or Github of preregistration for scientific studies.
The best preregistration plans will typically include a declared sample or population to observe (http://datacolada.org/64), or at least clear cut criteria for which participants or observations you will exclude.
I think for the type of economics/finance research I’m most familiar with, you often implicitly announce your sample when securing funding for a research proposal. E.g. if I’m trying to see if pursuing a momentum strategy with S&P 500 stocks is profitable (a la AQR’s work), it’s pretty obvious what the sample ought to be. This is partly why that meta study I linked to earlier was able to sniff out potential signs of p-hacking.
glenra · 2017-05-15
> There's been a few studies where identical resumes have been sent out with "white-sounding" and "non-white-sounding" names, with the white names getting more follow-up.
IIRC, that study finding didn't survive the replication crisis - the most recent resume studies didn't find bias against "non-white-sounding" names.
An optimistic interpretation of the discrepancy is that the earlier studies had picked low-class "black-sounding" names. It's hard to do this kind of study cleanly since names don't merely signify race, they also signify other things like socioeconomic status and age so to do it right you'd need to carefully validate that the specific names used don't differ in implications other than race; I don't think that has really been done yet.
(more here: http://datacolada.org/51 )
pocketsand · 2023-07-11
Check out the datacolada web site. https://datacolada.org/
AnthonyMouse · 2025-04-19
> E.g. how does this argument account for the data we have that someone with black sounding name will get less opportunity than someone with a white sounding name and an identical resume?
You're referring to a decades-old study that failed to replicate:
https://datacolada.org/51
(This is extremely common in social sciences.)
pyrale · 2024-03-15
The inquiry exposing the fraud is really interesting [1]. The bits about excel forensics really leave no space for doubt in that case.
[1]: https://datacolada.org/109
suddenclarity · 2024-03-15
I haven't followed this case but I found the blog posts doing the initial investigation quite interesting. Not only due to all the errors in the data but also because it gives some insight into how Excel works behind the scenes:
https://datacolada.org/109
https://datacolada.org/110
https://datacolada.org/111
https://datacolada.org/112
https://datacolada.org/114
I'm confused by her suing and her argument in court though. Is her argument that they didn't catch her in front of the computer so someone else might have fauxed the data or is her argument that no data was altered? Because it seems undeniable that the data was altered to align with the hypothesis.
faeriechangling · 2024-03-11
I'll hold my breath until I see this result replicated a few times using different methods.
This reminds me a lot of https://datacolada.org/98 and if I think about it, the data would be trivial to fake.
JohnKemeny · 2024-02-03
All scientists should sign on top of their papers that what they write is truthful.
See also https://datacolada.org/98
Sabinus · 2024-01-08
The first blog post of the Data Coladia fraud exposure series: https://datacolada.org/109
Francesca Gino's response: https://www.linkedin.com/posts/francescagino_i-want-to-be-very-clear-i…
empathy_m · 2023-07-26
This looks like an awesome toolchain to try to use for developing qualitative data sets in disciplines where the norm is to cut corners (e.g. https://datacolada.org/98).
Could you sell a version of this product where you promise to keep the buyers anonymous and confidential (and destroy the prompt and outputs after delivery)?
letmeinhere · 2023-07-14
Yep, there's new evidence that one of Ariely's frequent (honesty research) collaborators (Gino) was falsifying their data too https://datacolada.org/109
twobitshifter · 2023-07-14
One datacollada discovery (Dan Ariely sign at top/bottom) hinges on original cells in Calibri and forged cells in Cambria.
https://datacolada.org/98
manvillej · 2023-07-10
the fourth part of the data colada research has been posted: http://datacolada.org/112
dtjb · 2023-07-10
If you read the underlying findings [1], there were two behavioral ethics studies with fabricated results. The first was conducted in 2012 and was supervised by Ariely.
[1] https://datacolada.org/109
blamazon · 2023-07-10
Folks seeking a more neutral description of the situation may desire to click into the 4-part data-driven report that is the original source of the claims, which is linked in the article but also below:
https://datacolada.org/109
jbullock35 · 2023-06-22
The article in question has five co-authors. It reports four studies. Data Colada post 109 is about Study 1, which Gino handled.
An earlier Data Colada post -- https://datacolada.org/98 -- is about Study 3 in the same article. It seems that Ariely is the only co-author who might have handled the data for that study.
civilized · 2023-06-22
The evidence is available in the linked blog article from Data Colada, which is the real source of the story.
https://datacolada.org/109
https://datacolada.org/110
jamessb · 2023-06-21
The FT article says: 'A group of academics who compile the Data Colada blog about the evidence behind behavioural science has started publishing a series of posts in which they say they will detail “evidence of fraud in four academic papers” co-authored by Gino'.
The first two of these blog posts are:
https://datacolada.org/109
https://datacolada.org/110
resolutebat · 2023-06-21
The actual blog posts make it crystal clear that the data was manipulated to make it provide the desired results.
https://datacolada.org/109
https://datacolada.org/110