Biologist decided to disavow his own study
science.org
science.org
He reported the academic dishonesty (copies of figures from another paper presented as new experimental results), and the university went to great lengths to keep the whole thing quiet.
5 years later, the lead researcher who faked the results still works at the university, the paper has not been retracted, and my friend left what became a pretty hostile lab (and AFAIK is still under NDA)
At that point, what is left aside from trying to publicly remove your own name from a paper you know is fraudulent?
see the Alzheimer's amyloid plaque cartel- https://www.statnews.com/2019/06/25/alzheimers-cabal-thwarte...
https://www.upi.com/Health_News/2019/12/31/Amyloid-plaques-m...
“It is difficult to get a man to understand something, when his salary depends on his not understanding it.”
Sounds like the entire history of academia (and most human endeavors, honestly)
> We spend drastically more government money on research funding than in the past
You didn't define "in the past" but pure science spending has generally declined as a percentage of GDP and has barely changed in inflation adjusted dollars in the last twenty years.
> yet get less in return*
As measured how?
Which stat? The poster didn't provide any.
> In actual dollars not inflation adjusted dollars.
Scientists are paid salaries and need to eat and stuff, so it doesn't really make sense to me to ignore inflation there.
We'll still ultimately learn more about what's true in our universe, but it will take even longer and be more expensive than necessary.
The new thing would have to have different incentives, or the same problems would just re-occur.
Some examples of where this is being done: open-access journals, study pre-registration, requirement to publish source code. No one's saying the old way is banned now, it exists in parallel with all those innovations. Some will stick, others will fizzle out.
... then see your new system get attacked from all angles by the existing establishment. Smears on TV, 'fact-checkers' in social networks, bans and cancellations, etc.
The stakeholders of the existing system you are trying to replace won't go quietly into the night.
In particular, out of the 695 amino acids that constitute the APP protein, all the mutations that cause disease are located within the 42 amino acids that comprise the amyloid peptide, or just next to it.
It’s definitely not normal, and should actually raise some red flags. I would like to know which institution does this sort of things.
I was intimidated into signed an NDA which says I can't talk about the NDA. The threats were serious, and I don't talk about that NDA.
Unless you're at senior levels of your university, you wouldn't know about sketchy stuff like this going on.
The way to solve this would be to extend restrictions on federal grant funding and financial aid to only encompass institutions which don't use tools like this.
When the supply of scientists far exceeds the demand, there must be a way to decide who gets jobs and funding. Some people favor objective measurements, because they are less prone to corruption and nepotism than subjective ones. But when you have objective measurements, people will try to game them, and some will use dishonest means.
- You need to be available for contracting work, but many people prefer to go from one full-time job to another. Or they might not want to leave their current job, unless they get a much better offer.
- You need to be in the right location. Relocating without a job offer is risky.
- Especially strong candidates are likely to get many offers when they go looking, which is a good position to be in for salary negotiations. But work-as-interview arrangements make it hard to interview with more than one company at a time.
- Anything resembling visa support is probably off the table.
This plus a very easy coding question (think fizzbuzz) works really well at a small enough company that you can ensure that all interviewers:
1. Are far enough above the hiring bar that they can comfortably tell when other people are (people near the bar have a harder time distinguishing whether someone is a bit above or below the bar)
2. Are invested enough in the company to want to do a good job interviewing
3. Are thoughtful enough to gauge the competencies the company needs for the role. Those needs are not universal, I know developers who thrive in super technical no-nonsense roles who flounder in product-y work.
If a company is very large or growing rapidly it is nearly impossible to guarantee these things. My experience is that companies that don't adopt something like the industry norms lose control of their hiring bar. In some cases this is fine (sometimes you need quantity, not quality) but if the business relied on that high technical bar there's generally a cascading effect of initiatives failing, good engineers leaving, and more initiatives failing because of it. I've seen this tank a few startups in the 50-100 engineer rage.
Absolute fav :)
Best guesses I've got for how to combat the situation: - Have a diversity of metrics that are somewhat independent, somewhat anti-correlated (if I'm using the right words) - Have non-measurable goals that you refer back to
Anyone else got ideas for hedging against the failure mode of Goodhart's Law? :)
I have seen this in a company where they would change employee objectives every year, and also promote and demote people into unexpected new roles. This seemed to result in everyone understanding that next year things can be very different, and does not give people enough time to corrupt a metric. So everyone needs to build good working relations with superiors and subordinates, and not only try to meet objectives but also do well for the company in general, because next year roles can be reversed.
This system is tiring for everyone though...
https://www.science.org/content/article/this-scientist-accus...
It appears that sequences were neither uploaded to Genbank nor to the BOLD system at the time of the study. It does appear that sequences related to this article have been uploaded to Genbank in September 2020, but as this happened six years after publication and no voucher information is presented in the article, post-publication review was unable to confirm whether these sequences were indeed derived from the research described in this article.
https://link.springer.com/article/10.1007%2Fs10531-021-02316...
So many of our institutions have been corrupted from the inside by the more ruthless nature of careerism that’s taken over the modern world. Bond rating agencies endorse junk debt, doctors say opiates aren’t addictive any more, epidemiologists say the politics of your protest determine if it’s safe during a pandemic, elite educational institutions tell you going into crippling debt will be good for you.
Literally every institution, including ones supposed to serve the public, is rotting from the inside out from people bullshitting to keep their job or get promoted.
The strong incentives to cheat is very new in academia. I've seen academia for about a quarter-century, and at least MIT went from a pretty honest place to a shark tank full of crooks.
That's not a general "the world was once better," but a very specific "academic fraud has grown out-of-control in the past 25 years."
On the other hand, if your name is on an article, you must do your homework and know if what you are signing is realistic or not.
Definitely not the case when you're an undergraduate trying to get your foot in the door with your first paper.
In the US, afaik, the purpose of undergraduate work is to learn the material of the field. A Phd program then teaches the student how to do research in a field and the student goes on to do actual research only then.
An undergraduate would do a research paper as independent study or a special department class I think. With the provision that advisor is doing a lot of hand holding.
So handing fake data to someone's whose effectively a complete newby and letting them process it and publish it as their first achievement, is really despicable as the article seems to describe.
And there are plenty of cases where you won't notice anything is off unless you actually examine the raw data in detail, which isn't really realistic as it's an enormous amount of effort and duplicated work. There are certainly cases where co-authors should notice that something is off about the data. But there are plenty of situations where there is no reason at all to suspect a problem with the data, and where there is no way to know that someone is simply providing false data.
“Don’t put your name on something you don’t understand” is a good rule to live by. All the authors share responsibility. We cannot accept credit when all works well and reject blame when it does not.
> But there are plenty of situations where there is no reason at all to suspect a problem with the data, and where there is no way to know that someone is simply providing false data.
Publishing an article with better results is definitely the best thing to do then. Short of that, publishing a letter to the editor sharing concerns is a good way of putting your doubts on record. Just removing your name looks too much like washing your hands of the whole thing.
The original paper was published while the first author was an undergraduate, btw, so they probably didn't have a complete grasp of the issue.
I wonder if she'd fund researchers to attempt to replicate crucial publications / studies.
I expect that she has enough money to put a serious dent in this problem, at least if she focused on some specific area.
In this case, Gates would be the funding agency. And its published findings would hopefully steer much academic teaching and research away from bad assumptions.
The Hellinga scandal at Duke is very typical. This is what is looks like from outside: https://www.buffalo.edu/content/dam/www/news/imported/pdf/Ma...
And this is an inside account: https://caffeinatedchemist.medium.com/hellinga-66449cf4699e
It's exclusively an accountability issue.
Science is hard. Replicating an experiment is costly in time and money.
The papers in scientific field are organized like a tree. A good paper by a good and credible researcher will be a base node that other papers branch off of. A poor or not credible paper will be leaf that won't be cited by others. Testing all the poor papers would be tremendous waste of time. And it wouldn't catch fraudsters as such because not being replicable isn't proof of fraud.
Most rich people are less interested in funding replication studies than they are discoveries, but sometimes they do invest heavily in infrastructure development like Chan-Zuckerberg.
Science can't survive if the answer to faking claims is to replicate every study. The reason is that replicating a study done in complete seriousness and honesty is hard and is not guaranteed to yield the same result. So "can't be replicated" shouldn't be a black mark - "was basically made up" should be a black mark.
And also, if faking stuff becomes standard with replication the main guard, an easy way to get points is announce the fake replication of a given study.
Yes, a study failing to replicate is not in itself evidence of fraud. Nor is it definitive evidence that the effect shown in the original doesn't exist, since a replication can be messed up too.
But science needs replication because we're ultimately trying to learn about the world, and an effect we can't reliably produce is usually an effect we're not sure exists. Replication sometimes catching dishonesty is a bonus: failure to replicate calls the effect into question, which calls the method used to find that effect into question, which sometimes calls into question whether that method was actually used. (Or in the case of bad statistics, whether it was chosen to dishonestly show a certain result.) It's the appropriate propagation of doubt.
It’s not a trial to determine the guilt of a supposed fraudster. This is what institutions should do because fraud is a breach of contract (in all sane institutions).
However, the main problem is made up results in the scientific record. This is what we attempt to solve by retracting articles. And from the point of view of the scientific record, there is no meaningful difference between a made up result and something that cannot be replicated.
I don't think "the scientific record" is more than a metaphor that's referred to in talking of retraction. Still, if one were to imagine a hypothetical scientific record, you would have to imagine something that's inherently malleable. Even if most basic results are never going to be disproved, so fraction of even the most basic result are going to be disproved.
The thing, it's unusual but not unheard-of for a researcher to come up with a result, have it not be replicated, have the result languish and then have a later researcher take advantage of the result, either show it could be replicated or otherwise explaining what was actually going on.
But if the explanation for a strange result is "they faked it", then the impact of the situation is different.
Moreover, non-reproducibility is a reason for retraction but I don't think it's the main reason and well know papers that haven't reproduced aren't necessarily retracted Afaik.
Which is to say, I think there's still an important difference between non-reproducible for unknown reasons and know-to-be-fabricated data.
Article on retraction practices and "correcting the scientific record" https://pubs.acs.org/doi/10.1021/acs.chemmater.9b00897
I wonder whether any dark humor was involved when making that initialism.