@erinhengel@albertobisin You can recreate the sus dataset. The mystery RA is supposed to have done that by accident (again, step 1 is a crucial assumption here) and completely by coincidence that string of extremely weird decisions by the RA led to perfectly hypothesis consistent data. 2/2
@erinhengel@albertobisin No it's not, strip away all the fluff and simplify what is claimed there;
1. If you reorganize the dataset in a specific way (why?)
2. Swap around 2 chunks of 9 columns & 8 rows
3. Then you copy 2 more blocks of data (no mention of size) and paste them somewhere else
1/2
@erinhengel@albertobisin -the RA got the data for 3b, this can be seen in the e-mail convo.
-ok, do you buy that argument? That someone could have accidentally swapped a lot of lines and cells that just so happen to be consistent with the hypothesis?
@erinhengel@albertobisin Ok so then what's the alternative posed: Only FG had access to the qualtrics, she downloaded the data to a USB stick, gave it to a different RA never mentioned, quickly that RA faked the data (knowing the predictions), gave the new file to FG and then everyone immediately forgot?
@erinhengel@andre_quentin@lessig@albertobisin person can ruin your life. I would not recommend a graduate student or junior faculty to do so. 2/2
(not saying anything until you are sure enough would take ~1 year you can't spend on your own research, would also not recommend that to a junior).
@erinhengel@andre_quentin@lessig@albertobisin As I said, in the case I am familiar enough with to know the co-authors, the vast majority did not have any suspicions, some did bring up suspicions but did not feel institutional support and moved on (as you see publicly and explicitly claiming fraud against a powerful .. 1/2
@erinhengel@albertobisin But the RA didn't have access to the data (only the similar study 3b), claims to have never used USB sticks and Gino herself claims that there is no way the RA could have done it.
(the 39% not used is the podcast being manipulative imo, why should every difference be used?)
@erinhengel@andre_quentin@lessig@albertobisin For the case I'm intimately familiar with, a lot of it was not knowing better, some of it was a lack of resources/support to handle such things within existing power dynamics, and maybe some rare motivated reasoning.
@erinhengel@albertobisin So you can check my claims:
summary: page 18&19
interview RA: page 415
Co-author interview: page 354
RA e-mail correspondence: page 723
"Bad actor" defense I didn't go into because its silly but happy to do so: page 668
report: https://t.co/PZLgeEMkiT
@erinhengel@albertobisin 2/2 Importantly, according to the case docs, neither the RA nor the co-authors had access to that Qualtrics, and the file was posted by Gino. There's no evidence of the RA ever e-mailed her a cleaned dataset, and the RA denies doing any data cleaning on that file.
@erinhengel@albertobisin Ok so that is one example where it is clear that the Lessig podcast only addresses some accusations but avoids discussing others.
I can give a lot more examples if you want but I don't want to come across like I'm trying to steamroll you ^^
@erinhengel@albertobisin You are not addressing my question - Why, in 2023, did Francesca Gino's account upload a datafile *made to look like it was created in 2010*, to replace the 2010 datafile that Harvard argued showed evidence of tampering?
@erinhengel@albertobisin Ok, what's the explanation for why in 2023 Gino uploaded a datafile which date was manipulated to look like a 2010 file to replace a 2010 file showing evidence that the data was manipulated?
@erinhengel@albertobisin The podcast is very good at creating a "vibe" that something was off during the investigation, but once you critically engage with any of the arguments they all kind of fall apart.
@peterevoss@GaryMarcus@hendrycks@Yoshua_Bengio Sure but excluding queen from the inference seems a mistake no? The observed moves are consistent with rook and queen. So why is the conclusion rook when queen is equally valid?
@peterevoss@GaryMarcus@hendrycks@Yoshua_Bengio Sincere q, in the vid the bishop learns new moves by observing a few horizontal and a few vertical moves and then starts moving like a rook. But moving horizontally and vertically is equally consistent with a rook as a queen. So why is behaving like a rook the "correct" response?