"For more than 150 trials, Carlisle got access to anonymized individual participant data (IPD). By studying the IPD spreadsheets, he judged that 44% of these trials contained at least some flawed data: impossible statistics, incorrect calculations or duplicated numbers or figures, for instance. And in 26% of the papers had problems that were so widespread that the trial was impossible to trust, he judged — either because the authors were incompetent, or because they had faked the data."
So, 70% fake/flawed. The finding falls in line with other large scale replication studies in medicine, which have had replication success rates ranging from 11 to 44%. [1] It's quite difficult to imagine why studies where a positive outcome is a gateway to billions of dollars in profits, while a negative outcome would result in substantial losses, might end up being somehow less than accurate.
"It is simply no longer possible to believe much of the clinical research that is published, or to rely on the judgment of trusted physicians or authoritative medical guidelines. I take no pleasure in this conclusion, which I reached slowly and reluctantly over my two decades as an editor of the New England Journal of Medicine." -Marcia Angell
"If the image of medicine I have conveyed is one wherein medicine lurches along, riven by internal professional power struggles, impelled this way and that by arbitrary economic and sociopolitical forces, and sustained by bodies of myth and rhetoric that are elaborated in response to major threats to its survival, then that is the image supported by this study." -Evelleen Richards, "Vitamin C and Cancer: Medicine or Politics?"
Serving as editor of a major journal in any discipline can make one cynical. I read many years ago a quote from a mathematics editor that a large fraction of published proofs were faulty, and from a computer science editor that a large fraction of published theorems were false. (My foggy memory gives 1/2 and 1/3 for the fractions, but I don't remember which was which.)
Ah, but don’t worry folks, those vaccines which generated $40B in profits for Pfizer and Moderna, those are just A-ok. Trust the science, which in this case is perfect, unassailable and uncorruptible (unlike everything else the pharma industry does). Don’t question the faith.
"Trust the science" is dumb, but "trust that the single-most reviewed medical studies of the past decade with 30k+ participants aren't outright bullshit" is enough for me. The risk for those studies is probably closer to what happened with Astrazeneca, where a very rare side effect got missed because 30k people weren't enough to detect it.
I think skepticism is healthy to a degree, but ironically, the more adversarial the general public is to an area of study the more rigorous it usually is.
Peer review is not a magic spell that reveals the truth. “Most-reviewed” is meaningless here if the underlying data was not collected or reported properly.
Peer review, no. Review by a substantial percentage of world-wide immunologists along with everyone who's even remotely interested and a bunch of people who believe your work is killing people, still not a magic spell, but it's a hell of a lot better.
Did they review the raw data or did they just read the paper? AFAIK, Pfizer did not release the "summaries" of the data until March 2022 (forced by the FDA if i recall), long after these debates reached critical mass. I don't believe they released the data itself still. Google makes this hard to search for because it's still censoring...
Yeah, because hundreds of health authorities simultaneously collecting their own data (often with larger groups of each cohort) wouldn't be able to replicate or disprove the outcome of the original research.
You on the other hand believe that, somehow, absolutely independent (and often confronted) authorities, both governmental and private, would magically agree on something false instead of calling each other out.
> "Trust the science" is dumb, but "trust that the single-most reviewed medical studies of the past decade with 30k+ participants aren't outright bullshit" is enough for me.
What does "aren't outright bullshit" translate into in quantitative (%) terms?
I'd say bullshit means data that's worthless, because it's fabricated, collected by P-hacking, or otherwise means nothing. There's no useful quantitative measurement that tells you more than "Professor Bullshit just filled out random junk in Excel."
I mean I’m as vaccinated as anyone, but the originally reported 95% effectiveness against symptomatic COVID-19 with the monovalent vaccine just doesn’t seem realistic in hindsight. Who knows what complex set of incentives was at play. If you’re running a trial site and your business relies on continued contracts from Pfizer, are you reporting every cough, fever, or headache?
I wonder if the 95% effectiveness was actually true: against the strains prevalent at the time they ran the trial. But effectiveness against later strains was diminished.
I don't think the claim was 95% against symptomatic covid, it was against hospitalization or death.
The vaccine was tested against the original virus but Omicron was almost a different disease with much more immune escape. We're lucky the vaccine held up as well as it did. This is also why the alternative doctors that wanted everyone to catch covid to build up herd immunity were wrong, but nobody seems to bring them up. These discussions just devolve into the usual polarized political talking points.
There is no need to think, you can go read it. There was no indication that it was any help against severe disease, hospitalization, or death. The touted 95% was indeed against some definition of “symptomatic Covid” (a subjective one). In the all cause mortality section, there were more deaths in the vaccine arm than in the placebo arm. (But that was not statistically significant). At any time before 2020, the study would have been laughed at. But scientific standards were thrown out the door in 2020, and have not yet returned. (And … as it turns out from this and similar studies e.g. Ioannidis’ seminal paper showing most research is wrong from 2005 - it wasn’t as good as most people thought in 2020 already)
Whether the original study said so or not, that is absolutely how it was universally sold to the public. This [1] is a collection of various high profile individuals talking about the efficacy of the vaccines, why that means you won't carry the virus, you won't get sick, and and how that will completely stop the spread of COVID.
It's only as this very obviously failed to be the case that the metric was completely shifted to hospitalization/death. I'd also add this is about the time that the 'public messaging' swapped from talking about efficacy and other topics to outright vitriol and attacks on unvaccinated individuals, which is probably where the politicization of the topic began.
By "the politics" you're including individuals like Fauci, the head of the CDC, and so on. These are individuals that were perceived as "the science." Beyond this, a peer comment motivated me to look up the original trial results and reporting. And yeah, it was entirely about cases. Here [1] are the data from Moderna. Oddly enough, it seems to have been moved, if not removed, from their website, but fortunately nothing ever dies on the internet.
"There were 11 COVID‑19 cases in the Moderna COVID‑19 Vaccine group and 185 cases in the placebo group, with a vaccine efficacy of 94.1% (95% confidence interval of 89.3% to 96.8%)."
They were claiming it outright prevented COVID, as vaccines generally do. So "the politics" and "the science" were in lockstep on this one. As were they when they seamlessly dropped this narrative and swapped over to hospitalization/death.
Severe side effects or death cases were excluded from the Pfizer phase 3 study by conveniently shifting blame on patients, inventing diagnoses or claiming they had a Covid infection.
Even if all that were not intentional, it would certainly raise the question why this should be the only case of fabricating data for studies. Since the RNA-vaccines are obviously not as positively effective as claimed, what else is there that might be not talked about?
And lastly why is the topic still that controversial when the safety and effectivity claims have not stood up to the real world test?
Early into the pandemic, the original COVID strains had like 40% rate of permanent aftereffects, and the hospitalisation and death rates were insane. At the time, the vaccines have definitely saved lives, and the "covered up severe side effects or death" you talk about never came up in the real world, so...
And those weren't simple studies they were self-policed and self-reported trials. Trials that determined FDA green lighting. How can the FDA operate like that when the incentives to profit are 40B x more than being honest.
As someone who works in Data Science at FAANG - if you look hard enough - there is something questionably wrong in every step of the data funnel.
And that's when I believe people do have a somewhat best effort to maximize profits. There are plenty of people that only care about career progression and think they can get away with lying and cheating their way to the top. They wouldn't believe that if it didn't work sometimes.
These medical studies are also run mainly to maximize profits, also by some career climbers. They are not run virtuously for the betterment of society.
So I would be astounded if they are as reliable as people might like to believe.
Maybe I'm just being grossly skeptical. Actually, I'd feel better if someone could convince me I'm completely unfounded here.
"And that's when I believe people do have a somewhat best effort to maximize profits."
Nope. I actually think that if you do scientific research as a company (profit) it may make you less bad/less likely to do fraud compared to academia (non-profit).
Reason is that there are more ways to punish you, employees, board, investors, etc in a profit seeking vehicle, and as a profit seeking vehicle being caught must be part of the profit seeking calculation – in the end, the world of reality/physics will weigh your contribution.
I believe there is evidence that there is more fraudulent scientific research happening in non-profit vehicles/academia. Take for example an area where there are fewer profit seeking companies participating - social sciences. It's dominated by academia. Now look at the replication rate of social sciences.
I used to work data entry as a research assistant, for some non-profit academia.
We saw flaws in the data collection - basically the people tasked to collect data were being lazy and some were making stuff up. We know the made up stuff when we see it. Outliers are fine and some groups do better than expected, but an entire group from one data collecter shouldn't be 100% outliers.
But we had to enter the data anyway. We were told to smooth the bad data to what was expected. So the outliers that were low were smoothed high, the high outliers were left alone because they seemed right. But those of us who were spending hundreds of hours on data entry had an intuitive feel of what an outlier looked like.
IMO everything should have been entered as is and the computer data would just be filtered out if it was deemed from a corrupt source. But the data in the computer was biased to match what the research wanted to prove.
So I agree that non-profits can be corrupt too, just because of the incentives each part of the way. We were being paid about half a cent per column of data. So some assistants were lazy and filling in data that could be right, or skimming on fields like address which are longer and less likely to be flagged.
I think it's true that industry is more likely to produce reliable research compared to academia, but for a different reason.
In academia, you essentially have the student who does the work, the professor, and the person who funded the grant, and that's essentially the sum total of people supervising the data. There's not a lot of people to call you out on fudged (or outright faked) data, and all of them are likely to be very invested in the success of the research.
Turn to industry, and you have a similar set of people--the worker, the manager, and the head of the research department--except maybe a few more levels of manager (depending on the scale of the project). But since the goal is usually productization in industrial research, you usually have to turn to the product divisions and convince their executive chain as well of the merits of your research. And unlike everybody else mentioned so far, this group of people isn't invested in the success of the research. You might even be competing against other research teams that have different alternatives, and those people are going to be actively invested in the failure of your research so that their research makes it instead.
Also the guys in the product team are interested in whether they can actually reproduce eg your novel synthesis method for your favourite molecule on a large scale.
The 'product' of academic research is a published paper. The product of industrial research is an actual product.
(This does not apply when your research is about eg effectiveness of a new drug. The product people can sell that drug on the strength of that research. Whether that research replicates or not is only of indirect concern in that case.)
Completely anecdotally, I find this to be true. I do pre-clinical research (disclaimer: for a pharma company), and if you hand me a paper describing some research result in my field, it’s much more likely to reproduce in my hands if it’s from industry. I know from talking to my academic research friends that there are a ton of shortcuts, rotating inexperienced staff, and misaligned incentives. I’m not saying pharma is perfect, but in my experience, they’re more reproducible.
Fwiw, I do not know of any data in my realm which have been molded, cherry picked, intentionally misrepresented, falsified, or otherwise fake or flawed. I don’t work with clinical trial data, so if that were happening, it wouldn’t be on my desk.
I used to work in agricultural R&D in the private sector, and we collaborated with several universities.
The biggest difference I saw was that universities were very short-term focused, while we were more long-term focused.
The universities had a constant churn of personnel, as new PhD candidates appeared and old ones left, whereas our own researchers and technicians stuck around for far longer.
Additionally, they were so hyperfocused on grants and papers that they tended to not put as much effort into replication, since that didn't pay their bills. By comparison, we typically repeated our experiments ad nauseum; it was common for us to perform the same experiment twice a year (once in the northern hemisphere and then again in the southern hemisphere) for a decade or more, gradually iterating and refining our processes along the way.
Even if we initially got negative results, we'd beat that dead horse for a few years to make sure. Occasionally it turned out not to be so dead after all.
So many people have the opinion that private research must be flawed because of the profit motive, but the profit motive ensures that someone will be motivated to take oversight seriously, and have the power to punish misbehavior.
Free markets, as ugly as they are sometimes, are still the best way we have of ensuring that incentives align with outcomes.
War also works, to a certain extent, as a source of truth to align incentives with outcomes. But it's horribly expensive even in terms of economics alone, not to mention the human tragedy.
Luckily free markets work as a backstop, too. Add in free movement of people (who often want to come to better run places), and free movement of capital, and you have a winning combination.
Another stroke of luck: even if you only implement a very partial version of 'free', you still get partial benefits. Slightly freer markets are typically slightly more efficient. It's not an all or nothing proposition.
But work in social sciences will often be inherently harder to replicate due to, for example, groups of human beings being more complex than groups of proteins - so there are a lot more variables, confounding factors, difficulties doing double-blind randomised trials, and so on...
Yes, but that's why the companies like Google or Facebook spend so much time and effort on their social science experiments.
Granted, they are mostly interested in a very small sliver of social science: 'how can you get people to directly or indirectly spend more time online and look at more ads'; but they are very, very interested in getting robust results that replicate well. They are also interested in figuring out how the results vary between different cultures and over time.
You can fake everything except a well designed A/B test. At FAANG scale, a statistically significant A/B test requirement will stop the worst fraud before it hits the user.
Seriously though, as a person who has built related systems at FAANG, yes this problem exists there. Your beautiful cathedral of an A/B testing framework is covered in knobs that are just perfect for p-hacking.
>Actually, I'd feel better if someone could convince me I'm completely unfounded here.
Unfortunately, your skepticism is not unfounded. Those in the industry conclude the same. Take, for instance, the editor in chief of The Lancet:
>The case against science is straightforward: much of the scientific literature, perhaps half, may simply be untrue. Afflicted by studies with small sample sizes, tiny effects, invalid exploratory analyses, and flagrant conflicts of interest, together with an obsession for pursuing fashionable trends of dubious importance, science has taken a turn towards darkness.
I know someone who’s fabricated almost all their work. Their lab leads always cover it up and don’t retract their work. The crazy part is this person has failed upwards and is currently a senior researcher at a major research institution. They’ve been at UCLA and UCSF for medical research.
I’ve literally never seen data-driven business decisions that weren’t using fatally flawed datasets or methods so bad that you’d be a fool to believe you were getting anything but gibberish out of them, except in trivial cases.
You quickly learn not to be the guy pointing out the problem that means we’ll need several people and months or years to gather and analyze data that would allow them to (maybe) support or disprove their conclusion, though. Nobody wants to hear it… because they don’t actually care, they just want to present themselves as doing data-driven decision making, for reasons of ego or for (personal, or company) marketing. It’s all gut feelings and big personalities pushing companies this way and that, once you cut through the pretend-science shit.
“Yeah, that graph looks great (soul dies a little) let’s do it”
Objectivity and honesty can be hard to find if all someone cares about is their reputation as a competent researcher or climbing the ladder. What do you think a potential solution would be for this? I feel like even in my own experience, trying something out and it not working feels like failure, when in fact to proclaim it a success or "fix" it is truly what harms both the endeavor for truth and the people reliant on the outcomes of these surveys
> Objectivity and honesty can be hard to find if all someone cares about is their reputation as a competent researcher or climbing the ladder.
It seems like destroying the reputation and career of people who fake science would be a great start. If you're willing to fake data and lie to get results, there will always be an industry who'd love to hire you no matter how tarnished your reputation is. We need a better means to hold researchers accountable and we need to stop putting any amount of faith in any research that hasn't been independently verified through replication.
Today the lobby for orange juice manufactures can pay a scientist to fake research which shows that drinking orange juice makes you more attractive, and then pay publications to broadcast that headline to the world to increase sales. We should have some means to hold publications responsible for this as well.
> destroying the reputation and career of people who fake science would be a great start
When so many reports are faulty and fraudulent, that might instead be the great start of destroying the careers of those who would have revealed the fraudulent research?
I wonder what'd happen if researchers got compensated and funding based on other things, unrelated to papers published. But what would that be
> I wonder what'd happen if researchers got compensated and funding based on other things, unrelated to papers published. But what would that be
See what they do in the parts of industry where they need their research to work.
Eg how do battery manufacturers compensate and incentives their researchers that are aiming to improve various characteristics of batteries? How do steel mills manage and reward their metallurgists? How does Intel's research work?
Aha, so they get paid the same, regardless of if their research works or not. And possibly some bonuses if things go well for the company? And if they all the time do nothing useful and manipulate data, they get fired?
But that's different from a PhD student -- they're not embedded in any organization that would notice if the research works or not?
Maybe if the universities partnered somehow with different companies, and the researchers got extra compensation if a company decided to make real world use of the research?
(On top of some base salary)
But who would determine if a company had made use of a certain research paper? What would the company gain, by keeping track and reporting back? Maybe more good research
(but I'd guess few companies would be that much forward-looking?)
> Aha, so they get paid the same, regardless of if their research works or not.
I don't know. Is that speculation on your part, or something you figured out?
> But who would determine if a company had made use of a certain research paper? What would the company gain, by keeping track and reporting back? Maybe more good research
> (but I'd guess few companies would be that much forward-looking?)
That's why I am saying we should look what real companies are actually doing already in reality. We might have to leave our armchairs for that.
> Is that speculation on your part, or something you figured out
That's just normal monthly wages, how things usually work.
> look what real companies are actually doing already
But you can't look at what companies are doing now, to find out if new research is useful? The companies can't yet have started doing the things that any new & good research enables (since it wasn't known before).
Could take years until they make use of the research
> Researchers at universities also do these kinds of things because it helps them advance their careers.
This is a huge problem and in my opinion is mostly due to bad incentive structures and bad statistical/methodological education. I'm sure there are plenty of cases where there is intentional or at least known malpractice, but I would argue that most bad research is done in good faith.
When I was working on a PhD in biostatistics with a focus on causal inference among other things, I frequently helped out friends in other departments with data analysis. More often than not, people were working with sample sizes that are too small to provide enough power to answer their questions, or questions that simply could not be answered by their study design. (e.g. answering causal questions from observational data*).
In once instance, a friend in an environmental science program had data from an experiment she conducted where she failed to find evidence to support her primary hypothesis. It's nearly impossible to publish null results, and she didn't have funding to collect more data and had to get a paper out of it.
She wound up doing textbook p-hacking; testing a ton of post-hoc hypotheses on subsets of data. I tried to reel things back but I couldn't convince her to not continue because "that's how they do things" in her field. In reality she didn't really have a choice if she wanted to make progress towards her degree. She was a very smart person, and p-hacking is conceptually not hard to understand, but she was incentivized to not understand it or to not look at her research in that way.
* Research in causal inference is mostly about rigorously defining the (untestable) causal assumptions you must make and developing methods to answer causal questions from observational data. Even if an argument can be made that you can make those assumptions in a particular case, there is another layer of modeling assumptions you'll end up making depending on the method you're using. In my experience it's pretty rare that you can really have much confidence that your conclusions about a causal question if you can't run a real experiment.
It’s so interesting to hear you say that. I became disillusioned with causal methods for observational data for similar reasons. You can’t often model your way to interesting inferences without an experiment.
[1]: "Evaluating Coca-Cola’s attempts to influence public health ‘in their own words’: analysis of Coca-Cola emails with public health academics leading the Global Energy Balance Network" https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10200649/
It is not all that difficult to select or find an interesting scientific questions for which a Yes, a No, or even a Maybe is publishable and interesting. Two on my highest impact studies have been fun “No” studies—-
1. No, there is minimal or no numerical matching between populations of neurons in retina (ganglion cells) and populations of principal neurons in their CNS target (the thalamus). That demolished the plausible/attractive numerical matching hypothesis. I was trying valiantly to support it ;-)
2. No, there is no strong coupling of volumes of different brain regions due to “developmental constraints” in brain growth patterns.
https://pubmed.ncbi.nlm.nih.gov/23011133/
That idea just struck me as silly from an evolutionary and comparative perspective. We were happy to call it into doubt.
I suspect many of the comments are being made by damn fine programmers who know right from wrong ;-) a la Dijkstra. But in biology and clinical research, defining right and wrong is an ill-defined problem with lots of barely tangible and invisible confounders.
We should still demand well designed, implemented, and analyzed experimental or observational data sets.
However, that alone is not nearly enough to ensure meaningful and generalizable results. The meta-analyses were supposed to help at this level for clinical trials but have been gamed by bad actors with career objective that don’t consider patient outcomes even a bit.
Highlighting the problem is a huge step forward and it looks like AI may provide some near-future help along with more complete data release requirements.
If you have done biology—-
Hot. Wet. Mess. But beautiful.
> That doesn't mean you're a bad scientist, just an unlucky one. But it does mean you can't get tenure.
That sounds like a really easy problem to solve. Just treat valid science as important regardless of the results. The results shouldn't matter unless they've been replicated and verified anyway.
We should reward quality work, not simply the number of research papers (since it's easy to churn out trash) or what the results are (because until they are verified they could be faked).
Peer review is a joke. Too often it's a rubber stamp because there's no accountability for journals that fail to do the job. Unless peer review means something, the standard should change so that published papers only count if they're independently replicated and verified.
Which is why you fake your data into looking successful, and you don’t even go through the effort of publishing the failed studies (leading to its own additional problems in understanding what’s true and what we know)
I think you misread the article slightly. The 26% of zombie papers are inside the 44% of papers that contained at least some flawed data. Look at the figure just below this quote where the blue bar, indicating papers he thought were ok, covers more than 50%.
Which, ironically, just shows how easy it is to make mistakes not just in complicated statistical methods, but in basic interpretation of what numbers mean in the first place.
not to mention that everyone here forgets to mention that the majority of submitted papers with suspected fraud came from countries with known issues with fraudulent science publishing. e.g. China.
The USA, for example, looked much better (although not perfect) at 6% zombie.
I don't think the statement reads that the 44% and the 26% should be additive. Especially given the zombie graphic where it looks like they overlap the 26 on top of the 44, where the orange bar is the 26% and the remaining yellow bar is the 44%
It's more complex than just that. Sure, there's the people trying to make a dollar who are willing to do bad science in order to get the result they want. But there's also the general publication bias against replication studies - who wants to read them, and who wants to do them (they're not usually seen as prestigious academically: most academics want to test their ideas, not those of others.
And then there's cultural differences in which people sometimes see a negative result as a "failure", don't publish it as a result, and instead skew the data and lie their asses off in order to gain prestige in their career. As long as nobody double checks you, you're good.
> ut there's also the general publication bias against replication studies - who wants to read them, and who wants to do them (they're not usually seen as prestigious academically: most academics want to test their ideas, not those of others.
Academia seems like the idea place for this. Why not require a certain number of replicated studies in order to get a degree? Universities could then be constantly churning out replication studies.
More importantly, why do we bother taking anything that hasn't been replicated seriously? Anyone who publishes a paper that hasn't been verified shouldn't get any kind of meaningful recognition or "credit" for their discovery until it's been independently confirmed.
Since anyone can publish trash, having your work validated should be the only means of gaining prestige in your career.
It's worse. First, there's a selection bias in what trials the author can get data for. Second, the 44% + 26% are just those where the author can detect problems from reading their data alone. If someone convincingly fakes their data, the author can't detect that.
> The finding falls in line with other large scale replication studies in medicine, which have had replication success rates ranging from 11 to 44%.
These numbers seem almost wholly unrelated. A perfectly good study may be extremely difficult to replicate (or even the original purpose of replication - the experiment as described in the paper may simply not be sufficient); and an attempt at replication (or refutation), successful or not, is under the same pressure to be faked or flawed as the original paper.
What I wonder is, since it is so easy for this one researcher to catch these flaws and lies, why are these not being caught by peer review?
Increasing seems like peer review is a waste of everyone's time, and it's be better to just publishing when you think your ready on the likes of arXiv, and let everyone examine and criticize it.
Run it up the flagpole, see if it gets saluted or shot down. Seems like there's be much more of an incentive to run confirmation studies.
Why? Because its just cargo cult science at the end of the day. Once you incentivize clinical trials to be used as a tool to unlock massive profits through an arbitrary process with a single regulator as judge it WILL be hacked. At every possible part of the process including regulators.
> Carlisle got access to anonymized individual participant data (IPD)
I'm not in the industry so my question might have an obvious answer to those of you who are: How would one go about getting IPD if you wanted to run your own analysis of trial data or other data-driven research?
You'll need to reach out to the study authors with a request. If they are interested (you're going to publish something noteworthy with a citation for them (low chance), you want to bring them in on some funded research (better chance), etc) then they'll push it to their Institutional Review Board (a group of usually faculty and sometimes administrative staff at a University/Hospital/Org) who will review the request, the conditions of the initial data collection, legal restrictions, and then decide if they'll proceed with setting up some sort of IRB agreement / data use agreement. Unless you're a tenured professor somewhere or a respected researcher with some outside group then you probably won't get past any of those steps. Even allegedly anonymized data comes with the risk of exposure (and real penalties) not to mention the administrative overhead (expense, time, attention) that you'll need to be able to cover the cost of through some funded research. That research, btw, will also need to be through some IRB structure. You can tap a private firm that acts as an IRB but that's another process entirely and most certainly requires fat stacks of cash. Legal privacy concerns, ethical concerns, careerism (nobody wants you to find the 'carry the two' you forgot so you can crash their career prospects), bloated expenses (somebody has to pay for all of that paperwork, all those IRB salaries, etc) and etc, etc, etc all keep reproducibility of individual data frozen. Even within the same institution. Within the same team! You have to tread lightly with reproduction.
I also see similar findings in tech where most experiment results are "fudged". In some cases, people run the "same" experiment 5+ times until one is stat sig.
I read those two percentages as not being added to each other. In other words, the 26% is a shittier subset of the 44%. At least that's what I think they mean.
While we're on the subject, how many studies period are faked or flawed to the point of being useless? It seems to me that the scientific community's reaction to the replication crisis has been to ignore it.
Sociology/Psych/Economics are almost all junk. Their conclusions may or may not be correct.
Medical studies are mostly junk. There's way too much financial incentive to show marginal improvement. Theraflu and anti-depressents come to mind. Both show a small effect in studies and launched billion dollar businesses.
Hard science stuff tends to be pretty good. Mostly just outright fraud and they usually end up getting caught.
Chemistry still has a ton of issues with people exaggerating numbers that can't be contested like yield and purity of reactions even among the top journals. Straight up fraud is rarer yes.
I used to think like that. Soft science = dubious, hard science = ok. Not any more though. Just recently there was LK99 in materials science though, which was pretty much the textbook definition of an exciting but non-replicable claim.
Also a lot of stuff is tough to classify. Are epidemiology or climatology "hard" sciences?
> While we're on the subject, how many studies period are faked or flawed to the point of being useless?
Or how many studies are useless, period? It's like publishing a memoir to Amazon. You can now say "author" on your resume, or when you're introduced or at cocktail parties but nobody finds any value in what you have to say. You can also use ChatGPT because people might not notice.
There is always value in expanding the breadth and depth of human knowledge, even if it doesn't seem useful to you, right now. That of course assumes that knowledge is true, which is the crux of the problem now.
Unfortunately, the preponderance of very low value research in the literature puts a significant burden on the scientists who have to sift through a lot of garbage to find what they're looking for. Even if the work is ostensibly correct (much of it is not), it really doesn't do anyone much good, except for the authors of course. But now every undergraduate and every parent's vanity project at Andover wants a first author contrib., so here we are.
I think language models will help us sift through the garbage. I have a hard time finding answers using Google (too much SEO garbage), but ChatGPT does a pretty good job of filtering the noise. It's not perfect, but it saves lots of time vs. Google.
> It seems to me that the scientific community's reaction to the replication crisis has been to ignore it.
They've always known so there hasn't actually been any new information from which to spur action. In the academic circles I've run in there has always been a strong mistrust of reported results and procedures based on past difficulties with internal efforts replicating results. Basically a right of passage for a grad student to be tasked with replicating work from an impossible paper.
That's the correct answer, but what is baffling is that this is news to so many people frequenting this website. Most everyone posting here has been though a university, but hardly anyone has been involved in the pursuit of research.
The replication crisis is even a thing because of very strong career incentives.
You might think that publishing about the replication crisis itself would be great for your career, but perhaps not. Maybe the incentives to be able to bullshit your way to a professorship are so great that no one wants to rock the boat.
Talking with a few students in science programs this seems to be the case, even when few higher ups want to admit it.
Proving something is false is less valued than proving it right. It's silly because if we valued the quality of science it should be the exact opposite.
PIs may not be able to raise money for replication studies. Some of this is a consequence of guidelines for federal funds to prevent waste on duplication of effort.
This points to the real problem and also where the responsibility and interest ought to be to fix it.
There’s no replication crisis for academics because they have a meatspace social network of academics; they go to conferences together and know each other. You can just ignore a paper if you know the author is an idiot.
If medical studies are faked, is it a problem? Presumable regulatory agencies are using these studies or something, right? Looks like the FDA and NSF need to fund some more replication studies.
At least in pre-clinical research, there’s also a mechanism by which regulatory agencies can sometimes force industry to pay for replication studies. It’s a little complicated, because sometimes there aren’t many groups who can reliably do the study, and clinical trials are so expensive and time-consuming to run, it would be difficult to do at that scale. Rodent studies are generally under $1M and monkey studies are generally under $5M (up a lot recently). Clinical trials are in the hundreds of millions of dollars and take years to run.
This goes far beyond there not being enough replication studies, the problem is that when replication studies are done the results don't reproduce because the results were bogus. https://en.wikipedia.org/wiki/Replication_crisis
If the replication studies were done at a reasonable rate there would be no incentive to produce bogus results because you'd be caught before you could go through an entire career as a "successful scientist."
There is no being caught :( Who is going to do the catching, exactly? There's no science police. Universities don't seem to care about anything except plagiarism.
Many papers aren't even "looks legit but doesn't replicate", they are self-evidently wrong just from reading them or the associated data. Just peer review done properly would solve that, but the problem is widespread because there's no real incentive to rock the boat. Once a bad technique or invalid approach gets through to publication a few times it becomes a new standard within the field because it lets people publish more papers.
I think the most natural place for reform is from the people who are paying for the research, if they realized that what they are spending money for has no value if it's being made up. Although there are a lot of scientists who by the principle of self-interest seem willing to waste their life in exchange for a small salary it makes less sense why the NSF would take that offer.
But that can't happen at the moment. The core problem is that civil servants and politicians are easily bullied by claims they aren't experts and thus can't judge if "expert" work is any good or not. So they just dish out money to anyone who works at a university, expecting them to self-police, but they don't.
There are only two ways to fix that, as far as I can see:
1. Have highly technical, scientifically trained and skeptical politicians who police academia and ensure research money is well spent.
2. Stop governments funding research in the first place.
(1) just doesn't seem feasible. There are so many problems there. (2) is feasible. So I suspect you're going to start seeing calls from the right to defund universities over the next ten years or so. The underlying motivation may be that universities are strongholds of the left, but the stated justification will be the high levels of research fraud.
Or working from historical analogies, 3) find scientists with a public and obvious track record of replication, and maybe even a few of the fraud hunters, and put them in charge of the NSF.
You could try to build a science police but it won't work. Congress tried that in the 70s with the ORI. After a few failed prosecutions they basically gave up and are now forgotten - still there and they mount a few prosecutions a year but clearly it has no effect. The sheer scale of the problem cannot be fixed by trying to find people breaking "the rules" because there are no rules to begin with.
Attempting to fix this within the framework of government would be incredibly distracting for the state, as it'd lead to endless fights and debates. For example, a reasonable person appointed to run the NSF might conclude they should just defund whole subfields of the social sciences, because there's no way to make them scientific at all, but good luck lasting in the job if you do that. A big part of the problem is that scientists are in hoc to a specific political ideology, so they will have powerful friends in Washington who want to see them stick around producing useful propaganda.
You don't need a science police - allocating one half of all funding to replicating existing experiments will solve most of the problem. Why would one scientist commit fraud to protect the reputation of someone they've never met?
Oh boy. They do. They do it all the time. Too many of them see themselves as part of a "team", defending each other against outsiders. The Slack chats between virologists talking about the COVID lab leak hypothesis are a good example of this, they were very explicit about their aims there.
Huge funding for replication studies is commonly suggested, but it won't work. I'm basing that view on experience of having read and reported invalid papers across several different fields, spent years following this story and its various ins and outs, and have written extensively about the problems with pseudo-science coming out of official institutions.
Replication is the polite society way to talk about a whole range of problems. People say there's a replication crisis because it sounds a lot better than saying there's a large scale fraud and incompetence crisis. But it's not like there are papers out there that abstractly don't replicate, nobody could have known, and when someone tries and fails then there's suddenly a whole process that's followed to root cause why it failed and fix it. There isn't anything even close to that. What actually happens is that very obvious problems are ignored or kicked into the long grass as long as possible, often indefinitely, and sometimes scientists even argue they're being victimized by those reporting problems.
This article is a reasonable place to start, if you're new to the problem space:
It raises points that are often overlooked due to the "replication crisis" framing, points like:
• Many papers that can be replicated are only replicable because they're of little value (e.g. poor children have worse exam results). We don't want to be flooded with highly replicable science that tells us only what everyone already knew.
• Many papers are replicable but their claims are still invalid because their logic or methodology is wrong. This category covers significant amounts of COVID science, for example.
• It's often unclear exactly what the definition of replicable is.
But zooming out from replicability for a moment, I wrote an article with about some of the more egregious problems here:
We're talking about a scientific culture that can't even stop obviously machine-generated text from being published on a massive scale. No other part of society has problems like this. Even talking about replication seems like a distraction whilst you have papers being published every day that contain obviously Photoshopped images, mathematically impossible numbers and AI generated text, yet nobody cares and getting even one case "resolved" (paper retracted) requires months or years of external pressure.
Think about how big the trust problems are with journalism. How to fix trust in journalism is talked about a lot on the conference circuit, but it's a hard problem. And yet newspapers don't publish garbled gibberish and faked images on a daily basis! That's the scale of the challenge faced with science reform.
And even if you somehow manage to drain that swamp, the first step before full replication is attempted is to get the original data, so the analysis steps can be replicated using the original numbers. Whole swathes of science fail here because academics refuse to reveal their data. This happens even when they've been made to sign a statement agreeing that they'll provide it on request. Again, their employers do nothing, so where's the pressure point? Only massive fiscal punishment could cause culture change here but the NSF has a single institutional goal of dishing out as much money as possible. To quote Alvaro de Menard,
Why is the Replication Markets project funded by the Department of Defense? If you look at the NSF's 2019 Performance Highlights, you'll find items such as "Foster a culture of inclusion through change management efforts" (Status: "Achieved") and "Inform applicants whether their proposals have been declined or recommended for funding in a timely manner" (Status: "Not Achieved"). Pusillanimous reports repeat tired clichés about "training", "transparency", and a "culture of openness" while downplaying the scale of the problem and ignoring the incentives. No serious actions have followed from their recommendations.
It's not that they're trying and failing—they appear to be completely oblivious. We're talking about an organization with an 8 billion dollar budget that is responsible for a huge part of social science funding, and they can't manage to inform people that their grant was declined! These are the people we must depend on to fix everything.
If a university was actually in the business of seeking truth, there is much they could do to solve this.
* working groups "red teams" whose whole job it is to find weaknesses in papers published at the university
* post mortems after finding papers with serious flaws exposing the problem and coming up with constructive corrective actions
* funding / forcing researchers to devote a certain amount of time to replicating significant results
* working groups of experts in statistics and study design available to consult
* systems to register studies, methodologies, data sets, etc. with the aim of eliminating error and preventing post-hoc fishing expeditions for results
The whole-ass purpose of a university is seeking knowledge. They are fully capable of doing a better job of it but they don't because what they actually focus on are things like fundraising, professional sports teams, constructing attractive buildings, and advancing ranking.
Most universities would be better off just firing and not replacing 90% of their administration.
Universities can't do that, because they lack the expertise. Most of the time, the only people in a university capable of judging a study are the people who did it.
That's because universities are in the business of teaching. Apart from a few rare exceptions, universities don't have the money to hire redundant people. Instead of hiring many experts in the same topic, they prefer hiring a wider range of expertise, in order to provide better learning opportunities for the students.
> Universities can't do that, because they lack the expertise. Most of the time, the only people in a university capable of judging a study are the people who did it.
Statisticians can find faults in many studies. Universities even employ statisticians to act as consultants for research in other departments, as that raises the quality of the institution's publications.
Statisticians can find faults in many studies, but without a sufficient background in the topic, they can rarely tell if those faults actually matter. And the statisticians who have the background are usually already involved in the study anyway.
Maybe brush up on undergraduate statistics and experimental design so that they don't execute and publish obviously flawed studies? Perhaps they could apply such basic knowledge to the peer review process which is supposed to catch these things. They could stop defending their peers who publish bogus research, stop teaching about obviously flawed nonsense like the Stanford Prison "Experiment" and hold "scientists" like Philip Zimbardo in contempt. It's like asking how individual cops can make things better, of course they can't fix systematic problems but that doesn't mean there's nothing they can do.
It doesn’t have to be large scale. It can start with a single journal, with a single university, with a single scientist.
Make peer review public. Weight replication studies more. Make conducted peer reviews and replications an important metric for tenure and grants. Publish data and code alongside papers.
Some of those things were tried already. Some peer reviews are public, I've read them. Data and code are sometimes published too.
It doesn't help. It just means outsiders get to read bad peer reviews and bad code, but the right people within the system don't actually care. There's no way for the public to hold anyone to account, science is a law unto itself because politicians are generally not scientifically minded. They can be easily bullied with technobabble. The few that aren't susceptible to this don't tend to rise to the top.
There's literally thousands of different things they could do.
But why do anything when the real business of academia is in the tax-free hedge funds they operate and the government-subsidized international student gravy train? There's no short-term incentive to change anything.
I'm not nearly qualified to make this argument, but has anyone ever suggested that we collectively do away with the principle that one must publish original research in order to receive a PhD? Maybe in something like entomology, there are enough undescribed beetle species out there to supply myriads of dissertations, but in other fields it seems like you are just incentivizing trivial, useless, or fraudulent research.
IMO the problem might be more that in general we’ve got this perception that there’s a sort of “academic ranking,” or something like that, that puts:
Bachelors < Masters < PhD
Which, of course, not not really the case. A PhD in <field> is a specialization in creating novel research in <field>. In terms of actually applying <field>, a Masters ought to be as prestigious or whatever as a PhD. That it isn’t thought of that way seems to indicate, I dunno, maybe we need a new type of super-masters degree (one that gives you a cool title I guess).
Or, this will get me killed by some academics, but let’s just align with with the general public seems to think anyway: make a super-masters degree, give it the Dr title, make it the thing that indicates total mastery of a specific field (which is what the general public’s favorite Doctors, MDs, have, anyway) (to the extent to which a degree can even indicate that sort of thing, which is to say, not really, but it is as good as we’ve got). Then when PhDs can have a new title, Philosopher of <field>, haha.
My impression has been that in the sciences, there are a significant number of people who get into a graduate program, just do what they are told by their adviser for a few years, get handed a PhD, land a job on the recommendation of their adviser, and spend the rest of their careers elaborating on what they did as a student.
This is not necessarily bad, as there is a lot of drudge work that needs doing in science. The problem is the pressure to over-promote everything you do. The reality is that very few people, even among PhD's, have the capacity to do really original work. There was a study many years ago on the physics community that found that less than 10% of physicists published even one thing in their career that someone else found worth citing.
As a completely unqualified laymen my initial question would be "If you don't reward novelty why would anyone focused on career building want to be novel?" Not that I doubt academia has people who want to push the cutting edge forward (if anything that seems to be the only reason to go into academia vs. private industry usually) but if I was a fresh-faced PhD aspirant I'd want to take the most reliable route to getting my degree and treading ground someone else has already walked seems like a much safer way to do that than novel research if the reward at the end is the same.
But maybe that's a good thing? I can't actually say a reason I think it'd be that terrible except for the profs doing novel research that would lose their some of their student workforce.
That wouldn't fix anything at all. You still get that PhD to do research and publish. The only fix for this is funded and career-advancing randomized replication. We should lean harder into the scientific method, not withdraw from it. Absolutely nothing else will work.
Yes I work with these IPD spreadsheets every day and they all have typos often in the time/date column (issues with midnight and new years are very common) this can result in an erroneous reporting of something like average drug exposure or clearance if 1 or 2 subjects have uncaught errors in the reported data
I would expect 26% or more studies have these flaws but faked data is a different thing entirely and 26% fake would be incredibly worrying
How many studies are faked and how many just have a 'selective outcome'? Almost every nutritional study I read is at some point sponsored by a group that would benefit from a positive (or negative) outcome. I imagine several studies are run and only the one where the conditions led to a desired outcome are actually published. The researchers may know the results are inconsistent and there is an error somewhere, but it's not in their best interest to find the error and correct it.
True that, but the article does distinguish between "flawed" (44%) and "problems that were so widespread that the trial was impossible to trust" (26%).
Really, though, how/why would we expect otherwise? There's nothing in the system to prevent it, and plenty to incentivize it. There's really no good reason to expect (under the current system) that it would not happen (a lot). Without systemic change, it will not get better.
How do you discern between a study that is flawed and one that is faked? For instance, you could use some methodology that you know to be flawed, and allow that flawed methodology to bias your result in a pre-determined direction. If you're ever caught, you just claim it was an honest error, but it functions exactly the same as if you generated the data wholesale.
He groups them together because ultimately the result is that the science can't be trusted. He doesn't go so far to claim that one was intentionally faked vs gross incompetence.
My interpretation is that "flawed" study usually have honest data but wrong interpretation/analysis or wrong data gathering design so to make the conclusion essentially worthless. "Faked" study just fucking lie and may just provide rigged data to support a flawless conclusion.
They mean flawed to the point of being useless in a way that indicates incompetence or negligence or fraud. Not flawed like they had a typo or a negative result or didn't use MLA formatting in their citation.
Could you please stop posting unsubstantive comments and flamebait? You've unfortunately been doing it repeatedly. It's not what this site is for, and destroys what it is for.
My PhD advisors wife's job at a big pharma company was in a department that attempted to reproduce interesting papers. She claimed they only had a 25% success rate (this was early-mid 2000s)
When I first grasped how many fake papers are floating around, I gnashed my teeth and called for heinous penalties upon the fraudsters.
After a moment, I thought better, and decided that we should actually offer incentive for fraudulent papers! Liars will always have incentive to lie. When their lies are accepted by those who ought to be more skeptical, it's an indictment of the system, which should be more robust in detecting fraud.
I'm not as worried about faked/flawed studies as I am about pharmaceutical companies knowingly selling bad drugs to make billions of dollars with no recourse from the FDA. I'll never look at Big Pharma the same way again after watching Painkiller:
I'm sorry to say that I've personally witnessed some people in the next lab commit fraud. It was investigated and someone was fired, but I can believe that it is all too common.
Somebody tried to submit a paper with my name on it that was fraudulent (they took my final draft and goosed the numbers). I immediately balked and they withdrew it from submission. However, the scientist still got tenure and even retired in good standing. Sigh.
Isn't this an issue with the reviews rather than the publication attempts? Wouldn't the normal (if there is such a thing) approach between people be to credit a reviewer with the quality of their reviews or lack thereof? Basically, who signed off on it?
That is my understanding. Papers can be wrong for different reasons, some nefarious, some honest mistakes. The problem I wouldn't say is with the paper itself. It more of people citing it with a lack of peer review or lack of replication. And also peer review being like a stereotypical "LGTM" approval on a pull request.
We all know it is super-easy to cheat in science, but not much can be done about it. Short of requiring replication for every result published which isn't feasible if the study was costly to produce. And the problem isn't confined to medicine either. How many studies in hpc of the type "we setup this benchmark and our novel algorithm/implementation won!" aren't also faked or flawed?
The "setup this benchmark" happens in medicine all of the time, and it's super insidious and no one sees it. It's hard to tease out.
For example, say you have a new (patentable) drug that you're trying to get approved and replace an old (generic, cheap) drug. You need to prove that the new drug is at least as safe as the old one and/or more effective than the old one. Which sounds reasonable.
Let's say that the old drug is super safe and that the new one may not be. What you can do is set up your RCT so that you surreptitiously underdose the control arm so the safe drug looks less effective. And then you can say that your new drug is more effective than the old one.
Peer review doesn't notice this because you can easily hide it in the narrative of the methods section. I've seen it a couple of times.
So you can easily have "gold standard RCTs" that get the result you want just by subtle changes in the study design.
Plenty can be done about it. We can start by increasing the standards for experimental proof past words on a piece of paper.
Stream or record your experiment. If a 12 year old has the ability to stream his gaming life on twitch, scientists should be able to record what they are doing in more detail.
You could start a new journal with higher level of prestige that only publishes experiments that adhere to more modern methods of proof.
This seems overly simplistic, I mean part of the reason so many studies are so difficult is because they can happen over the period of months or years. This isn't like your High School chemistry experiment. These studies can be massive longitudinal studies that take years to conclude and work on.
If the answers were easy I'm sure someone would've implemented it already, it turns out those kind of things are hard.
I'm interested more in why it specifically would be hard to just show people what you are doing (or did) instead of telling them a story at the end of months or years.
Critical details and oversights can be lost in translation between reality and LaTeX that could be easily pointed out by third party observers.
Who is going to watch thousands[0] of hours of video, most of which will show a scientist fiddling around with setscrews or hunting down a loose ground wire?
And what scientist is going to be happy with constant surveillance that's intended to be shared (so no picking your nose, losing your temper, or making mistakes)?
[0] Literally, thousands. I just finished a paper about a project that we started three years ago. This was the main project for two people, plus a tech: 2000 hrs/yr * 3 * 80% effort = 4800 hours of footage.
That does sound like a lot of footage, but you wouldn't watch every second just as you likely haven't read every word of every scientific paper you've skimmed or referenced.
The footage is there to support inquiries into how you did your experiment, just as you would skim a youtube video about repairing an appliance to get to the part you need to see.
And yeah, it would have to be a new level of prestige backing this way of documenting your methods. Obviously people prefer being trusted, but p-hacking and replication problems have proven that we can't have nice things.
> We can start by increasing the standards for experimental proof past words on a piece of paper.
That is overly cynical. We are already well past that standard.
We require words on a piece of paper that were written by somebody from a recognizable institution, or by an AI operated by somebody from a recognizable institution.
If the study is costly to produce, then it's even more important to replicate it, otherwise you'd wasting a large amount of money on a study with no real sense of whether it is flawed or not.
I've read that the real reason why researches generated by war crimes are worthless, is because we can't replicate it. Using a result that its validity can't be checked is just a way to disaster. And this is researches with the highest cost (human lives) and we still can't use it.
Lots of "interesting" psychology experiments in the days of yore turns out to have lots of damaging confounding variables and we can't just redo the experiment because, well, we shouldn't.
It helps a lot but rests on the assumption that the people with whom you're pre-registering will check that you did the same thing you claimed you'd do. And also, that they won't tolerate gamed pre-registrations e.g. pre-registrations so vague that they admit many different post-hoc analyses.
Sometimes this is true. One example is when the people verifying the pre-registration work for a regulator. The introduction of pre-registration seems to be at least partly behind the collapse in pharma productivity:
17 of 30 studies (57%) published prior to 2000 showed a significant benefit of intervention on the primary outcome in comparison to only 2 among the 25 (8%) trials published after 2000 (χ2=12.2,df= 1, p=0.0005). There has been no change in the proportion of trials that compared treatment to placebo versus active comparator. Industry co-sponsorship was unrelated to the probability of reporting a significant benefit. Pre-registration in clinicaltrials.gov was strongly associated with the trend toward null findings.
But when journals do this the results seem to be worse. They aren't strongly incentivized to improve anything, so you can get studies that claim to be pre-registered but do something different to what the pre-registration said they'd do.
Much science is corrupted by financial conflicts of interest, politics where conclusion is already formed, chasing a higher number of citations, and experimental errors. We have a lot of data, even documentaries, on drug studies. Does anyone have collections of links on...
1. Evidence of widespread error and fraud in science in general?
2. Evaluations of many sub-fields of science like we do for drug studies showing how true or false they are?
How many replies will originate from those who actually read the linked content, engage with it thoughtfully, and sidestep their own biases, distortions, prejudices, and ignorance?
An even lower percentage.
This species and its cultures are suffering an accelerating cognitive landslide.
Coupled with the many other accelerating descents, it's a shame, really.
The few times I visited a hospital I met people there with complications from previous drugs. One with liver failure. Thanks to that experience no covid shot for me, plus I actually read the pfzier study the judge forced them to release.
I now view drug companies on the level of Cartels in Mexico with politicians in their pockets.
Sorry, I'm no longer convinced of this. I think there's a large range of procedures/treatments being done that don't work, that kill people all the time. And in the end you are often just trading short term risk, for a longer payoff.
Say you have prostate cancer. Should you risk the low chance of dying of prostate cancer 10 years from now, or the higher chance of incontinence, and medical complications from the treatment you undertake now.
That is why doctors are supposed to consent their patients and weigh those risks. For what it’s worth, the risk of incontinence form prostate cancer surgery has been dropping quickly as traditional surgery is replaced by robotic surgery, sparing the nerve that runs close under the prostate.
Pfizer was contracted to deliver a vaccine that works that is safe. Instead defrauded taxpayers for tens of billions of dollars, and delivered a medication with pages worth of side effects that has higher risk profile than covid for almost anyone under 50.
This topic concerns me a lot, since the Ivermectine phenomenum plus the Covid Vaccines. People mentioned sample sizes in several comments. What concerns me is that sample size requirements versus incidence of a problem (ex: covid deaths) is a very non-intuitive discussion for the general public, and even sometimes for people in technical fields.
Here is a simple tool that can help exploring sample sizes:
You know how (to use a familiar field) most of software development has become going through the motions, churning massive insecure bulk, doing nonsense rituals and making up myths about process and productivity, hopping jobs frequently, cutting corners and breaking rules, etc., and the purpose is usually not to produce trustworthy solutions, but only to further career and livelihood/wealth?
With everything we've been seeing in recent years on HN, about science reproducibility and fraud, and the complaints about commonplace fudging and fraud that you might hear privately from talking with PhDs/students in various fields... I wonder whether science has developed a similar alignment problem.
How many people in science careers are doing trustworthy science? And when they aren't, why not?
> You know how (to use a familiar field) most of software development has become going through the motions, churning massive insecure bulk, doing nonsense rituals and making up myths about process and productivity, hopping jobs frequently, cutting corners and breaking rules, etc., and the purpose is usually not to produce trustworthy solutions, but only to further career and livelihood/wealth?
Well furthering wealth is the reason why tech companies exist. The incentives in academia are completely different. You might be right, but i see no reason to expect similar behavior across such drastically different situations.
At least I get eggs from a neighbor and not “big egg”; I keep thinking about getting my own chickens but that’s not the way I want to feed Mr. Fox who lives in my neighborhood too.
My free range chickens are protected by free range fox predators that are communal with them. Sometimes they will eat a chicken or two, but mostly they let the chickens be because they keep the free range parasites in check.
I definitely count on stinging insects that live in holes in my house (or that make holes) to keep other stinging insects away.
What I've been told is that opossums are much worse than foxes in the sense that a fox will usually eat a chicken or two to survive but opossums seem to freak out and will kill all the hens in a henhouse in one go.
I've often wished I could talk with my cats but more than ever I wish I could ask them what they knew about the fox. There is this lady
The animals we're mostly concerned about at our farm are horses and cats. I don't think there is anything left in North America that can trouble a horse, unless you count mosquito-transmitted infectious diseases. I hear foxes are not dangerous to cats but I believe we've lost some to coyotes.
Mostly we've had people around, either tenants or neighbors, who keep chickens so we don't have to. I'll say the eggs from a small scale chicken operation taste a lot better than commercial eggs.
I definitely thought about trying to draw in the fox but as much as that British lady makes it look easy on Youtube Shorts the legends are that foxes can cause a lot of trouble.
I don't tell the SPCA, however, that I am getting a cat there because it's a replacement for one lost to predators. They die of old age, get hyperthyroidism, high blood pressure, become blind, walk around mindlessly and crash into things, and one morning you find them dead at the bottom of the stairs. I think they would find dying at the claws of a predator an honorable death.
As for kitsune they are the Japanese version of a myth that is widespread in East and South Asia, for instance Daji
and is sometimes believed to be the same entity. I was into anime for a long time (Urusei Yatsura was a watershed but it really goes back to seeing Star Blazers on TV) but lately I have been into Chinese pop culture like Three Kingdoms, Nezha and Wolf Warrior 2 and that's gotten me reading about Chinese mythology and one clear thing is a lot of Japanese mythology comes from China, for instance it seems there is a "world tree" in almost every JRPG that they climb to get to heaven and even if they call it Yggdrasil it is not from Norse mythology it is from Chinese mythology. (Turns out also a lot of Chinese mythology as well as neopagan ideas comes from India as well.)
Indeed, a klingon-like death for a cat would be honorable.
The Japanese do derive a lot of their culture from the Chinese, don't they? I would like to one day read some of the stuff around the Three Kingdoms. Alas! So much to read and do, and so little time!
What!? Everyone knows a glass of wine once a day is good for your heart, but also even a single glass a year is way too much and causes irreversible damage ;)
The old adage applies: "everything nice is either illegal or bad for your health". Or both, I would add.
Yes, of course. But have you read journalists reporting on science findings? It's always an extreme "scientists now claim one glass of wine a day is good for you!" and "scientists discovered that even one glass of wine will ruin your life forever".
Never the middle ground, it's always a shocking new finding "by science" (spoiler: scientists seldom say the things newspapers and pop-science/nutrition & health articles claim they say).
I read a lot of studies about gout, there's a lot of studies like "we found participants who consumed x cups of coffee lowered their uric acid" all of them follow the same pattern. They asked people to consume more liquids, and it lowered their uric acids.
Both of those are substances that should be harmful based on the effects of the major ingredients: e.g. caffeine is addictive and when I am deprived of it my schizotypy flares up and I have paranoid episodes (I yell at my wife “why the hell are you always standing where I want to go next?”) I have had two doctors tell me three other reasons why I should quit.
Look in the medical literature and it seems outright spammed by reports on the positive effects of caffeine and negative reports on any of the harmful effects one would expect.
Similarly the main active ingredient of red wine (alcohol) is harmful, red wine in particular causes a lot of discomfort, dispepsia, hangovers and other unpleasant effects if you get a bad vintage but look the literature and it is like it will transport you to a blue zone and you will love forever.
And you find those kind of papers spammed in “real” journals, not MDPI or “Frontiers” journals.
Yes but that's not something you can automatically draw inferences from. Exercise is harmful to you in a short enough time interval but benefits you on the long run.
Furan content in coffee too. Very low amount but still.
Coffee prevents headaches for me so I'll always drink it. And no it's not related to physical dependence although at this point the withdrawal will guarantee a headache.
Interesting. I was at one point diagnosed with "tension headaches" and prescribed a medication called Fioricet which is a combination of caffeine, acetaminophen, and a barbituate called butalbital. Incredibly effective. I'm sure the harm profile of coffee alone is lower so if it's totally eliminating headaches as a problem for you it's perhaps a better solution, but thought i'd throw that out there as it could be worth seeing a neurologist if the problem becomes unmanageable for you
I can relate. I never drank anything with caffeine and would get headaches fairly often. Headaches were never too bad or too often to need medical attention but was just normal part of life. I started drinking coffee on road trips and drives over 3 hours long and noticed that my headache coming on would go away right after. Now i drink coffee twice daily and i'll get a headache once a month at most.
I drank five or six cups of coffee a day for decades. I’d even have a cup before going to bed — that’s how tolerant I had become.
Got a mild flu/Covid/cold couple years ago. Better in a week. But, during the illness and since, the slightest bit of caffeine would make be incredibly wired to the point of panic attacks. Had to quit cold turkey. I’ve tried a cup now and again, and it’s the same thing: 6 hours of overwhelming anxiety.
Wierd. It’s like I became hypersensitive to caffeine. Oddly, though, nicotine doesn’t have that effect, and I always figured the two stimulants were similar.
Yerba mate is my caffeine of choice and I suggested it to my Dad who gets bad migraines. He claims that when he feels a migraine coming on, he can drink a can and it will result in a mild headache instead of forcing him to lie down in a dark room for hours.
... and I find it very hard to stay at a low dose. If I quit entirely for a few weeks I could probably manage one small coffee a day for a while but inevitably I'd have a rough night and then I need a small and then another small or maybe just a large and pretty soon I can be drinking two whole carafe a day.
I have the same issue w/ cannabis. Right now I have a few plants (legal) in the garden and also a bag that is going to a friend and I don't care. If I had a little puff though the next day I would want another little puff and another and in a week or so I would be like the guy in the Bob Marley song
It might be that those are common enough in shared datasets or when datasets are collected. So it is easy to draw interferences with them and various other measured factors.
I often chuckle that same people who vehemently defend vaccine safety studies also decry studies that show processed seed oils, glyphosate, aspartame, and hundreds of other molecules or compounds are unsafe over time. That's the beauty of medical studies! Whether they're good or bad, your own biases mean you can simply ignore the ones that don't align with your opinions.
Oh, before I forget! Red meat and saturated fat is terrible for you! Wait, no, it's actually sugar that's evil. Vegetables are great for you! Oh wait, most vegetables contain oxalates and other defense, mechanisms and compounds that are actually bad for you overtime.
This is all stated as if it is fact. It sounds believable to me, but how do you (and the rest of us) know that the sources you got this from aren't a part of the junk being called out by the article?
I dislike "the same people who" comments, but I agree with the sentiment. A lot of us have very little ability to determine the validity of this study over that, but will confidently voice an opinion anyway.
The only way (imo) to stay on firm ground is to acknowledge that someone published a thing saying xyz, and maybe that you are x% convinced by it. Can't get too far out over your skis going that route.
I think you're accidentally telling on yourself here. You're looking at somebody getting a result which is surprising to you but rather than being curious about how they might be on to something you're turning off your brain and assuming they're malfunctioning.
Something being in a category, such as "a study", doesn't tell you much about a thing. If you read multiple studies on vaccine safety critically and reason about them and what experts are saying about them, IMO most functional human being are going to reach the same general conclusion about vaccine safety. If you do the same thing on studies about seed oils or aspartame you're also going to come to the conclusion that they're safe! If you're not reaching these same results it doesn't necessary mean you're the one who is malfunctioning but you should seriously consider it and try again to learn what you might not know.
I've never met one of those people. In general people who decry vaccines (as a norm, not talking about the covid can of worms) tend to fall into the "alternative medicine" bucket and distrust all science studies, and those who trust vaccines tend to also trust other scientific studies...
Hmm -- in usual fashion I haven't read the original post here, but I'm guessing it didn't find that RCT trials that big pharma do are quite as useless as most 'medical studies' generally.
Do you think well-funded RCTs (like those that support vaccine safety) are just as weak as any old observational study?
But my question to the person saying it's problematic to defend vaccine studies and attack food results is: isn't it possible that you feel the research procedures used in one are superior to those used in another?
For example: vaccine safety study looks at 200,000 people and randomly assigns them to use or not use the vaccine. Coffee/red wine study looks at 30 people and surveys them about how they felt last week after drinking coffee/red wine. Looking at these two, I think it's fair to put more trust in the vaccine study.
So, 70% fake/flawed. The finding falls in line with other large scale replication studies in medicine, which have had replication success rates ranging from 11 to 44%. [1] It's quite difficult to imagine why studies where a positive outcome is a gateway to billions of dollars in profits, while a negative outcome would result in substantial losses, might end up being somehow less than accurate.
[1] - https://en.wikipedia.org/wiki/Replication_crisis#In_medicine
"It is simply no longer possible to believe much of the clinical research that is published, or to rely on the judgment of trusted physicians or authoritative medical guidelines. I take no pleasure in this conclusion, which I reached slowly and reluctantly over my two decades as an editor of the New England Journal of Medicine." -Marcia Angell
"If the image of medicine I have conveyed is one wherein medicine lurches along, riven by internal professional power struggles, impelled this way and that by arbitrary economic and sociopolitical forces, and sustained by bodies of myth and rhetoric that are elaborated in response to major threats to its survival, then that is the image supported by this study." -Evelleen Richards, "Vitamin C and Cancer: Medicine or Politics?"
I think skepticism is healthy to a degree, but ironically, the more adversarial the general public is to an area of study the more rigorous it usually is.
That's the-moon-landing-was-fake tier denial-ism.
You cannot compare the moon landing to the COVID circus.
What does "aren't outright bullshit" translate into in quantitative (%) terms?
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7745181/
Sample size was 43k people split in treatment and control group. Adding it here for context.
The vaccine was tested against the original virus but Omicron was almost a different disease with much more immune escape. We're lucky the vaccine held up as well as it did. This is also why the alternative doctors that wanted everyone to catch covid to build up herd immunity were wrong, but nobody seems to bring them up. These discussions just devolve into the usual polarized political talking points.
It's only as this very obviously failed to be the case that the metric was completely shifted to hospitalization/death. I'd also add this is about the time that the 'public messaging' swapped from talking about efficacy and other topics to outright vitriol and attacks on unvaccinated individuals, which is probably where the politicization of the topic began.
[1] - https://twitter.com/i/status/1472327161352798212
Trust the politics, blame the science?
"There were 11 COVID‑19 cases in the Moderna COVID‑19 Vaccine group and 185 cases in the placebo group, with a vaccine efficacy of 94.1% (95% confidence interval of 89.3% to 96.8%)."
They were claiming it outright prevented COVID, as vaccines generally do. So "the politics" and "the science" were in lockstep on this one. As were they when they seamlessly dropped this narrative and swapped over to hospitalization/death.
[1] - https://web.archive.org/web/20210202223626/https://www.moder...
Even if all that were not intentional, it would certainly raise the question why this should be the only case of fabricating data for studies. Since the RNA-vaccines are obviously not as positively effective as claimed, what else is there that might be not talked about? And lastly why is the topic still that controversial when the safety and effectivity claims have not stood up to the real world test?
File under: Follow the money
And that's when I believe people do have a somewhat best effort to maximize profits. There are plenty of people that only care about career progression and think they can get away with lying and cheating their way to the top. They wouldn't believe that if it didn't work sometimes.
These medical studies are also run mainly to maximize profits, also by some career climbers. They are not run virtuously for the betterment of society.
So I would be astounded if they are as reliable as people might like to believe.
Maybe I'm just being grossly skeptical. Actually, I'd feel better if someone could convince me I'm completely unfounded here.
Nope. I actually think that if you do scientific research as a company (profit) it may make you less bad/less likely to do fraud compared to academia (non-profit).
Reason is that there are more ways to punish you, employees, board, investors, etc in a profit seeking vehicle, and as a profit seeking vehicle being caught must be part of the profit seeking calculation – in the end, the world of reality/physics will weigh your contribution.
I believe there is evidence that there is more fraudulent scientific research happening in non-profit vehicles/academia. Take for example an area where there are fewer profit seeking companies participating - social sciences. It's dominated by academia. Now look at the replication rate of social sciences.
We saw flaws in the data collection - basically the people tasked to collect data were being lazy and some were making stuff up. We know the made up stuff when we see it. Outliers are fine and some groups do better than expected, but an entire group from one data collecter shouldn't be 100% outliers.
But we had to enter the data anyway. We were told to smooth the bad data to what was expected. So the outliers that were low were smoothed high, the high outliers were left alone because they seemed right. But those of us who were spending hundreds of hours on data entry had an intuitive feel of what an outlier looked like.
IMO everything should have been entered as is and the computer data would just be filtered out if it was deemed from a corrupt source. But the data in the computer was biased to match what the research wanted to prove.
So I agree that non-profits can be corrupt too, just because of the incentives each part of the way. We were being paid about half a cent per column of data. So some assistants were lazy and filling in data that could be right, or skimming on fields like address which are longer and less likely to be flagged.
In academia, you essentially have the student who does the work, the professor, and the person who funded the grant, and that's essentially the sum total of people supervising the data. There's not a lot of people to call you out on fudged (or outright faked) data, and all of them are likely to be very invested in the success of the research.
Turn to industry, and you have a similar set of people--the worker, the manager, and the head of the research department--except maybe a few more levels of manager (depending on the scale of the project). But since the goal is usually productization in industrial research, you usually have to turn to the product divisions and convince their executive chain as well of the merits of your research. And unlike everybody else mentioned so far, this group of people isn't invested in the success of the research. You might even be competing against other research teams that have different alternatives, and those people are going to be actively invested in the failure of your research so that their research makes it instead.
The 'product' of academic research is a published paper. The product of industrial research is an actual product.
(This does not apply when your research is about eg effectiveness of a new drug. The product people can sell that drug on the strength of that research. Whether that research replicates or not is only of indirect concern in that case.)
Fwiw, I do not know of any data in my realm which have been molded, cherry picked, intentionally misrepresented, falsified, or otherwise fake or flawed. I don’t work with clinical trial data, so if that were happening, it wouldn’t be on my desk.
The biggest difference I saw was that universities were very short-term focused, while we were more long-term focused.
The universities had a constant churn of personnel, as new PhD candidates appeared and old ones left, whereas our own researchers and technicians stuck around for far longer.
Additionally, they were so hyperfocused on grants and papers that they tended to not put as much effort into replication, since that didn't pay their bills. By comparison, we typically repeated our experiments ad nauseum; it was common for us to perform the same experiment twice a year (once in the northern hemisphere and then again in the southern hemisphere) for a decade or more, gradually iterating and refining our processes along the way.
Even if we initially got negative results, we'd beat that dead horse for a few years to make sure. Occasionally it turned out not to be so dead after all.
So many people have the opinion that private research must be flawed because of the profit motive, but the profit motive ensures that someone will be motivated to take oversight seriously, and have the power to punish misbehavior.
Free markets, as ugly as they are sometimes, are still the best way we have of ensuring that incentives align with outcomes.
War also works, to a certain extent, as a source of truth to align incentives with outcomes. But it's horribly expensive even in terms of economics alone, not to mention the human tragedy.
Luckily free markets work as a backstop, too. Add in free movement of people (who often want to come to better run places), and free movement of capital, and you have a winning combination.
Another stroke of luck: even if you only implement a very partial version of 'free', you still get partial benefits. Slightly freer markets are typically slightly more efficient. It's not an all or nothing proposition.
Granted, they are mostly interested in a very small sliver of social science: 'how can you get people to directly or indirectly spend more time online and look at more ads'; but they are very, very interested in getting robust results that replicate well. They are also interested in figuring out how the results vary between different cultures and over time.
You can fake everything except a well designed A/B test. At FAANG scale, a statistically significant A/B test requirement will stop the worst fraud before it hits the user.
Seriously though, as a person who has built related systems at FAANG, yes this problem exists there. Your beautiful cathedral of an A/B testing framework is covered in knobs that are just perfect for p-hacking.
Unfortunately, your skepticism is not unfounded. Those in the industry conclude the same. Take, for instance, the editor in chief of The Lancet:
>The case against science is straightforward: much of the scientific literature, perhaps half, may simply be untrue. Afflicted by studies with small sample sizes, tiny effects, invalid exploratory analyses, and flagrant conflicts of interest, together with an obsession for pursuing fashionable trends of dubious importance, science has taken a turn towards darkness.
https://www.thelancet.com/journals/lancet/article/PIIS0140-6...
Absolutely insane to watch.
You quickly learn not to be the guy pointing out the problem that means we’ll need several people and months or years to gather and analyze data that would allow them to (maybe) support or disprove their conclusion, though. Nobody wants to hear it… because they don’t actually care, they just want to present themselves as doing data-driven decision making, for reasons of ego or for (personal, or company) marketing. It’s all gut feelings and big personalities pushing companies this way and that, once you cut through the pretend-science shit.
“Yeah, that graph looks great (soul dies a little) let’s do it”
Oh, that hits home
It seems like destroying the reputation and career of people who fake science would be a great start. If you're willing to fake data and lie to get results, there will always be an industry who'd love to hire you no matter how tarnished your reputation is. We need a better means to hold researchers accountable and we need to stop putting any amount of faith in any research that hasn't been independently verified through replication.
Today the lobby for orange juice manufactures can pay a scientist to fake research which shows that drinking orange juice makes you more attractive, and then pay publications to broadcast that headline to the world to increase sales. We should have some means to hold publications responsible for this as well.
When so many reports are faulty and fraudulent, that might instead be the great start of destroying the careers of those who would have revealed the fraudulent research?
I wonder what'd happen if researchers got compensated and funding based on other things, unrelated to papers published. But what would that be
See what they do in the parts of industry where they need their research to work.
Eg how do battery manufacturers compensate and incentives their researchers that are aiming to improve various characteristics of batteries? How do steel mills manage and reward their metallurgists? How does Intel's research work?
But that's different from a PhD student -- they're not embedded in any organization that would notice if the research works or not?
Maybe if the universities partnered somehow with different companies, and the researchers got extra compensation if a company decided to make real world use of the research?
(On top of some base salary)
But who would determine if a company had made use of a certain research paper? What would the company gain, by keeping track and reporting back? Maybe more good research
(but I'd guess few companies would be that much forward-looking?)
I don't know. Is that speculation on your part, or something you figured out?
> But who would determine if a company had made use of a certain research paper? What would the company gain, by keeping track and reporting back? Maybe more good research
> (but I'd guess few companies would be that much forward-looking?)
That's why I am saying we should look what real companies are actually doing already in reality. We might have to leave our armchairs for that.
That's just normal monthly wages, how things usually work.
> look what real companies are actually doing already
But you can't look at what companies are doing now, to find out if new research is useful? The companies can't yet have started doing the things that any new & good research enables (since it wasn't known before).
Could take years until they make use of the research
(But maybe you meant something else)
Look at what processes worked (and didn't work!). Not at what specific inventions worked.
(Research by Google about psychological safety comes to my mind.)
I do think there is an even worse issue - which is funding. The money incentive means you can fund studies that support whatever you want.
This is a huge problem and in my opinion is mostly due to bad incentive structures and bad statistical/methodological education. I'm sure there are plenty of cases where there is intentional or at least known malpractice, but I would argue that most bad research is done in good faith.
When I was working on a PhD in biostatistics with a focus on causal inference among other things, I frequently helped out friends in other departments with data analysis. More often than not, people were working with sample sizes that are too small to provide enough power to answer their questions, or questions that simply could not be answered by their study design. (e.g. answering causal questions from observational data*).
In once instance, a friend in an environmental science program had data from an experiment she conducted where she failed to find evidence to support her primary hypothesis. It's nearly impossible to publish null results, and she didn't have funding to collect more data and had to get a paper out of it.
She wound up doing textbook p-hacking; testing a ton of post-hoc hypotheses on subsets of data. I tried to reel things back but I couldn't convince her to not continue because "that's how they do things" in her field. In reality she didn't really have a choice if she wanted to make progress towards her degree. She was a very smart person, and p-hacking is conceptually not hard to understand, but she was incentivized to not understand it or to not look at her research in that way.
* Research in causal inference is mostly about rigorously defining the (untestable) causal assumptions you must make and developing methods to answer causal questions from observational data. Even if an argument can be made that you can make those assumptions in a particular case, there is another layer of modeling assumptions you'll end up making depending on the method you're using. In my experience it's pretty rare that you can really have much confidence that your conclusions about a causal question if you can't run a real experiment.
Do you think CocaCola and the Sacklers had their own unique ideas shared by no one else? That we've filtered all scrupulous people out of industry?
Scruples are an abstraction at that scale.
[0]: Politico "Coca-Cola tried to influence CDC on research and policy, new report states" [https://www.politico.com/story/2019/01/29/coke-obesity-sugar...]
[1]: "Evaluating Coca-Cola’s attempts to influence public health ‘in their own words’: analysis of Coca-Cola emails with public health academics leading the Global Energy Balance Network" https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10200649/
[2]: Forbes: "Emails Reveal How Coca-Cola Shaped The Anti-Obesity Global Energy Balance Network" https://www.forbes.com/sites/nancyhuehnergarth/2015/11/24/em...
... into the multi-billion dollar companies GP is talking about.
That doesn't mean you're a bad scientist, just an unlucky one. But it does mean you can't get tenure.
So it's easy to understand why people fake results to secure a career.
1. No, there is minimal or no numerical matching between populations of neurons in retina (ganglion cells) and populations of principal neurons in their CNS target (the thalamus). That demolished the plausible/attractive numerical matching hypothesis. I was trying valiantly to support it ;-)
https://pubmed.ncbi.nlm.nih.gov/14657177/
2. No, there is no strong coupling of volumes of different brain regions due to “developmental constraints” in brain growth patterns. https://pubmed.ncbi.nlm.nih.gov/23011133/
That idea just struck me as silly from an evolutionary and comparative perspective. We were happy to call it into doubt.
I suspect many of the comments are being made by damn fine programmers who know right from wrong ;-) a la Dijkstra. But in biology and clinical research, defining right and wrong is an ill-defined problem with lots of barely tangible and invisible confounders.
We should still demand well designed, implemented, and analyzed experimental or observational data sets.
However, that alone is not nearly enough to ensure meaningful and generalizable results. The meta-analyses were supposed to help at this level for clinical trials but have been gamed by bad actors with career objective that don’t consider patient outcomes even a bit.
Highlighting the problem is a huge step forward and it looks like AI may provide some near-future help along with more complete data release requirements.
If you have done biology—- Hot. Wet. Mess. But beautiful.
That sounds like a really easy problem to solve. Just treat valid science as important regardless of the results. The results shouldn't matter unless they've been replicated and verified anyway.
Actually implementing it across the academic world seems much harder.
We should reward quality work, not simply the number of research papers (since it's easy to churn out trash) or what the results are (because until they are verified they could be faked).
I think you miss read that section? Only were 44% fake or flawed acording to this study.
The 26% that were very flawed is a subset of the 44% that were flawed in general. So those precentages should not be added together.
And then there's cultural differences in which people sometimes see a negative result as a "failure", don't publish it as a result, and instead skew the data and lie their asses off in order to gain prestige in their career. As long as nobody double checks you, you're good.
Academia seems like the idea place for this. Why not require a certain number of replicated studies in order to get a degree? Universities could then be constantly churning out replication studies.
More importantly, why do we bother taking anything that hasn't been replicated seriously? Anyone who publishes a paper that hasn't been verified shouldn't get any kind of meaningful recognition or "credit" for their discovery until it's been independently confirmed.
Since anyone can publish trash, having your work validated should be the only means of gaining prestige in your career.
Most of the studies that were problematic came from China and Egypt.
In other words, nothing new here.
[1] https://associationofanaesthetists-publications.onlinelibrar...
It's worse. First, there's a selection bias in what trials the author can get data for. Second, the 44% + 26% are just those where the author can detect problems from reading their data alone. If someone convincingly fakes their data, the author can't detect that.
These numbers seem almost wholly unrelated. A perfectly good study may be extremely difficult to replicate (or even the original purpose of replication - the experiment as described in the paper may simply not be sufficient); and an attempt at replication (or refutation), successful or not, is under the same pressure to be faked or flawed as the original paper.
Increasing seems like peer review is a waste of everyone's time, and it's be better to just publishing when you think your ready on the likes of arXiv, and let everyone examine and criticize it.
Run it up the flagpole, see if it gets saluted or shot down. Seems like there's be much more of an incentive to run confirmation studies.
Why? Because its just cargo cult science at the end of the day. Once you incentivize clinical trials to be used as a tool to unlock massive profits through an arbitrary process with a single regulator as judge it WILL be hacked. At every possible part of the process including regulators.
I'm not in the industry so my question might have an obvious answer to those of you who are: How would one go about getting IPD if you wanted to run your own analysis of trial data or other data-driven research?
Sociology/Psych/Economics are almost all junk. Their conclusions may or may not be correct.
Medical studies are mostly junk. There's way too much financial incentive to show marginal improvement. Theraflu and anti-depressents come to mind. Both show a small effect in studies and launched billion dollar businesses.
Hard science stuff tends to be pretty good. Mostly just outright fraud and they usually end up getting caught.
Also a lot of stuff is tough to classify. Are epidemiology or climatology "hard" sciences?
Or how many studies are useless, period? It's like publishing a memoir to Amazon. You can now say "author" on your resume, or when you're introduced or at cocktail parties but nobody finds any value in what you have to say. You can also use ChatGPT because people might not notice.
They've always known so there hasn't actually been any new information from which to spur action. In the academic circles I've run in there has always been a strong mistrust of reported results and procedures based on past difficulties with internal efforts replicating results. Basically a right of passage for a grad student to be tasked with replicating work from an impossible paper.
You might think that publishing about the replication crisis itself would be great for your career, but perhaps not. Maybe the incentives to be able to bullshit your way to a professorship are so great that no one wants to rock the boat.
Our whole economy is fueled by people bullshitting each other.
Proving something is false is less valued than proving it right. It's silly because if we valued the quality of science it should be the exact opposite.
There’s no replication crisis for academics because they have a meatspace social network of academics; they go to conferences together and know each other. You can just ignore a paper if you know the author is an idiot.
If medical studies are faked, is it a problem? Presumable regulatory agencies are using these studies or something, right? Looks like the FDA and NSF need to fund some more replication studies.
Many papers aren't even "looks legit but doesn't replicate", they are self-evidently wrong just from reading them or the associated data. Just peer review done properly would solve that, but the problem is widespread because there's no real incentive to rock the boat. Once a bad technique or invalid approach gets through to publication a few times it becomes a new standard within the field because it lets people publish more papers.
There are only two ways to fix that, as far as I can see:
1. Have highly technical, scientifically trained and skeptical politicians who police academia and ensure research money is well spent.
2. Stop governments funding research in the first place.
(1) just doesn't seem feasible. There are so many problems there. (2) is feasible. So I suspect you're going to start seeing calls from the right to defund universities over the next ten years or so. The underlying motivation may be that universities are strongholds of the left, but the stated justification will be the high levels of research fraud.
Attempting to fix this within the framework of government would be incredibly distracting for the state, as it'd lead to endless fights and debates. For example, a reasonable person appointed to run the NSF might conclude they should just defund whole subfields of the social sciences, because there's no way to make them scientific at all, but good luck lasting in the job if you do that. A big part of the problem is that scientists are in hoc to a specific political ideology, so they will have powerful friends in Washington who want to see them stick around producing useful propaganda.
Huge funding for replication studies is commonly suggested, but it won't work. I'm basing that view on experience of having read and reported invalid papers across several different fields, spent years following this story and its various ins and outs, and have written extensively about the problems with pseudo-science coming out of official institutions.
Replication is the polite society way to talk about a whole range of problems. People say there's a replication crisis because it sounds a lot better than saying there's a large scale fraud and incompetence crisis. But it's not like there are papers out there that abstractly don't replicate, nobody could have known, and when someone tries and fails then there's suddenly a whole process that's followed to root cause why it failed and fix it. There isn't anything even close to that. What actually happens is that very obvious problems are ignored or kicked into the long grass as long as possible, often indefinitely, and sometimes scientists even argue they're being victimized by those reporting problems.
This article is a reasonable place to start, if you're new to the problem space:
https://fantasticanachronism.com/2020/09/11/whats-wrong-with...
It raises points that are often overlooked due to the "replication crisis" framing, points like:
• Many papers that can be replicated are only replicable because they're of little value (e.g. poor children have worse exam results). We don't want to be flooded with highly replicable science that tells us only what everyone already knew.
• Many papers are replicable but their claims are still invalid because their logic or methodology is wrong. This category covers significant amounts of COVID science, for example.
• It's often unclear exactly what the definition of replicable is.
But zooming out from replicability for a moment, I wrote an article with about some of the more egregious problems here:
https://blog.plan99.net/fake-science-part-i-7e9764571422
We're talking about a scientific culture that can't even stop obviously machine-generated text from being published on a massive scale. No other part of society has problems like this. Even talking about replication seems like a distraction whilst you have papers being published every day that contain obviously Photoshopped images, mathematically impossible numbers and AI generated text, yet nobody cares and getting even one case "resolved" (paper retracted) requires months or years of external pressure.
Think about how big the trust problems are with journalism. How to fix trust in journalism is talked about a lot on the conference circuit, but it's a hard problem. And yet newspapers don't publish garbled gibberish and faked images on a daily basis! That's the scale of the challenge faced with science reform.
And even if you somehow manage to drain that swamp, the first step before full replication is attempted is to get the original data, so the analysis steps can be replicated using the original numbers. Whole swathes of science fail here because academics refuse to reveal their data. This happens even when they've been made to sign a statement agreeing that they'll provide it on request. Again, their employers do nothing, so where's the pressure point? Only massive fiscal punishment could cause culture change here but the NSF has a single institutional goal of dishing out as much money as possible. To quote Alvaro de Menard,
Why is the Replication Markets project funded by the Department of Defense? If you look at the NSF's 2019 Performance Highlights, you'll find items such as "Foster a culture of inclusion through change management efforts" (Status: "Achieved") and "Inform applicants whether their proposals have been declined or recommended for funding in a timely manner" (Status: "Not Achieved"). Pusillanimous reports repeat tired clichés about "training", "transparency", and a "culture of openness" while downplaying the scale of the problem and ignoring the incentives. No serious actions have followed from their recommendations.
It's not that they're trying and failing—they appear to be completely oblivious. We're talking about an organization with an 8 billion dollar budget that is responsible for a huge part of social science funding, and they can't manage to inform people that their grant was declined! These are the people we must depend on to fix everything.
What can they do? It's an incredibly hard problem to solve. It's like asking why the buisness community has done nothing to adress the housing crisis.
Large scale culture changes, or the entire structure of the way science is conducted and funded, would be the only solutions
* working groups "red teams" whose whole job it is to find weaknesses in papers published at the university
* post mortems after finding papers with serious flaws exposing the problem and coming up with constructive corrective actions
* funding / forcing researchers to devote a certain amount of time to replicating significant results
* working groups of experts in statistics and study design available to consult
* systems to register studies, methodologies, data sets, etc. with the aim of eliminating error and preventing post-hoc fishing expeditions for results
The whole-ass purpose of a university is seeking knowledge. They are fully capable of doing a better job of it but they don't because what they actually focus on are things like fundraising, professional sports teams, constructing attractive buildings, and advancing ranking.
Most universities would be better off just firing and not replacing 90% of their administration.
That's because universities are in the business of teaching. Apart from a few rare exceptions, universities don't have the money to hire redundant people. Instead of hiring many experts in the same topic, they prefer hiring a wider range of expertise, in order to provide better learning opportunities for the students.
Statisticians can find faults in many studies. Universities even employ statisticians to act as consultants for research in other departments, as that raises the quality of the institution's publications.
Make peer review public. Weight replication studies more. Make conducted peer reviews and replications an important metric for tenure and grants. Publish data and code alongside papers.
It doesn't help. It just means outsiders get to read bad peer reviews and bad code, but the right people within the system don't actually care. There's no way for the public to hold anyone to account, science is a law unto itself because politicians are generally not scientifically minded. They can be easily bullied with technobabble. The few that aren't susceptible to this don't tend to rise to the top.
There's literally thousands of different things they could do.
But why do anything when the real business of academia is in the tax-free hedge funds they operate and the government-subsidized international student gravy train? There's no short-term incentive to change anything.
Dude. The scientific community created the replication crisis.
We are not ignoring it, I promise you. The reaction is nuanced by no one who matters is ignoring it.
This is putting it very mildly.
Bachelors < Masters < PhD
Which, of course, not not really the case. A PhD in <field> is a specialization in creating novel research in <field>. In terms of actually applying <field>, a Masters ought to be as prestigious or whatever as a PhD. That it isn’t thought of that way seems to indicate, I dunno, maybe we need a new type of super-masters degree (one that gives you a cool title I guess).
Or, this will get me killed by some academics, but let’s just align with with the general public seems to think anyway: make a super-masters degree, give it the Dr title, make it the thing that indicates total mastery of a specific field (which is what the general public’s favorite Doctors, MDs, have, anyway) (to the extent to which a degree can even indicate that sort of thing, which is to say, not really, but it is as good as we’ve got). Then when PhDs can have a new title, Philosopher of <field>, haha.
This is not necessarily bad, as there is a lot of drudge work that needs doing in science. The problem is the pressure to over-promote everything you do. The reality is that very few people, even among PhD's, have the capacity to do really original work. There was a study many years ago on the physics community that found that less than 10% of physicists published even one thing in their career that someone else found worth citing.
But maybe that's a good thing? I can't actually say a reason I think it'd be that terrible except for the profs doing novel research that would lose their some of their student workforce.
I would expect 26% or more studies have these flaws but faked data is a different thing entirely and 26% fake would be incredibly worrying
Really, though, how/why would we expect otherwise? There's nothing in the system to prevent it, and plenty to incentivize it. There's really no good reason to expect (under the current system) that it would not happen (a lot). Without systemic change, it will not get better.
If you wouldn't mind reviewing https://news.ycombinator.com/newsguidelines.html and taking the intended spirit of the site more to heart, we'd be grateful.
After a moment, I thought better, and decided that we should actually offer incentive for fraudulent papers! Liars will always have incentive to lie. When their lies are accepted by those who ought to be more skeptical, it's an indictment of the system, which should be more robust in detecting fraud.
https://www.netflix.com/title/81095069
For example, say you have a new (patentable) drug that you're trying to get approved and replace an old (generic, cheap) drug. You need to prove that the new drug is at least as safe as the old one and/or more effective than the old one. Which sounds reasonable.
Let's say that the old drug is super safe and that the new one may not be. What you can do is set up your RCT so that you surreptitiously underdose the control arm so the safe drug looks less effective. And then you can say that your new drug is more effective than the old one.
Peer review doesn't notice this because you can easily hide it in the narrative of the methods section. I've seen it a couple of times.
So you can easily have "gold standard RCTs" that get the result you want just by subtle changes in the study design.
Stream or record your experiment. If a 12 year old has the ability to stream his gaming life on twitch, scientists should be able to record what they are doing in more detail.
You could start a new journal with higher level of prestige that only publishes experiments that adhere to more modern methods of proof.
If the answers were easy I'm sure someone would've implemented it already, it turns out those kind of things are hard.
I'd recommend checking out the following article. https://slatestarcodex.com/2014/04/28/the-control-group-is-o...
Critical details and oversights can be lost in translation between reality and LaTeX that could be easily pointed out by third party observers.
And what scientist is going to be happy with constant surveillance that's intended to be shared (so no picking your nose, losing your temper, or making mistakes)?
[0] Literally, thousands. I just finished a paper about a project that we started three years ago. This was the main project for two people, plus a tech: 2000 hrs/yr * 3 * 80% effort = 4800 hours of footage.
The footage is there to support inquiries into how you did your experiment, just as you would skim a youtube video about repairing an appliance to get to the part you need to see.
And yeah, it would have to be a new level of prestige backing this way of documenting your methods. Obviously people prefer being trusted, but p-hacking and replication problems have proven that we can't have nice things.
That is overly cynical. We are already well past that standard.
We require words on a piece of paper that were written by somebody from a recognizable institution, or by an AI operated by somebody from a recognizable institution.
Lots of "interesting" psychology experiments in the days of yore turns out to have lots of damaging confounding variables and we can't just redo the experiment because, well, we shouldn't.
Sometimes this is true. One example is when the people verifying the pre-registration work for a regulator. The introduction of pre-registration seems to be at least partly behind the collapse in pharma productivity:
https://journals.plos.org/plosone/article?id=10.1371/journal...
17 of 30 studies (57%) published prior to 2000 showed a significant benefit of intervention on the primary outcome in comparison to only 2 among the 25 (8%) trials published after 2000 (χ2=12.2,df= 1, p=0.0005). There has been no change in the proportion of trials that compared treatment to placebo versus active comparator. Industry co-sponsorship was unrelated to the probability of reporting a significant benefit. Pre-registration in clinicaltrials.gov was strongly associated with the trend toward null findings.
But when journals do this the results seem to be worse. They aren't strongly incentivized to improve anything, so you can get studies that claim to be pre-registered but do something different to what the pre-registration said they'd do.
1. Evidence of widespread error and fraud in science in general?
2. Evaluations of many sub-fields of science like we do for drug studies showing how true or false they are?
An even lower percentage.
This species and its cultures are suffering an accelerating cognitive landslide.
Coupled with the many other accelerating descents, it's a shame, really.
I think we almost made it.
Almost.
If published studies that have been replicated were the highest incentive instead, maybe it would reduce the risk of faked studies.
https://www.youtube.com/watch?v=z6IO2DZjOkY
Why couldn't a bad actor just fake the raw data? Isn't that what Climategate was all about?
Edit: See the Dan Ariely drama for an example.
"How many medical studies of medical studies are faked or flawed? A meta report on meta studies."
https://xkcd.com/2755/
Well… yeah? All medications have a list of possible side effects in the leaflet. It’s just that the disease tends to be worse.
Say you have prostate cancer. Should you risk the low chance of dying of prostate cancer 10 years from now, or the higher chance of incontinence, and medical complications from the treatment you undertake now.
Can you send a link to it?
Here is a simple tool that can help exploring sample sizes:
https://lo.gic.li/lgc
With everything we've been seeing in recent years on HN, about science reproducibility and fraud, and the complaints about commonplace fudging and fraud that you might hear privately from talking with PhDs/students in various fields... I wonder whether science has developed a similar alignment problem.
How many people in science careers are doing trustworthy science? And when they aren't, why not?
Well furthering wealth is the reason why tech companies exist. The incentives in academia are completely different. You might be right, but i see no reason to expect similar behavior across such drastically different situations.
Nothing is more natural and free range than the occasional murder between animals.
What I've been told is that opossums are much worse than foxes in the sense that a fox will usually eat a chicken or two to survive but opossums seem to freak out and will kill all the hens in a henhouse in one go.
I've often wished I could talk with my cats but more than ever I wish I could ask them what they knew about the fox. There is this lady
https://www.youtube.com/@debs3289
who meets them in the street, has them come to her door, and feeds them chicken (!) Secretly I imagine that the fox is really a
https://en.wikipedia.org/wiki/Kitsune
and my wife is always reminding me that it has just one tail, not nine.
Mostly we've had people around, either tenants or neighbors, who keep chickens so we don't have to. I'll say the eggs from a small scale chicken operation taste a lot better than commercial eggs.
I definitely thought about trying to draw in the fox but as much as that British lady makes it look easy on Youtube Shorts the legends are that foxes can cause a lot of trouble.
Am I out of place if I say I also like coyotes? (Remember: city boy, so nothing is at stake for me here. I also like wolves!)
And yeah, Japanese folklore has taught me that it's best to avoid kitsune. Though they sometimes turn into magical women who help you?
I don't tell the SPCA, however, that I am getting a cat there because it's a replacement for one lost to predators. They die of old age, get hyperthyroidism, high blood pressure, become blind, walk around mindlessly and crash into things, and one morning you find them dead at the bottom of the stairs. I think they would find dying at the claws of a predator an honorable death.
As for kitsune they are the Japanese version of a myth that is widespread in East and South Asia, for instance Daji
https://en.wikipedia.org/wiki/Daji
is probably isomorphic to Tamamo-no-mae
https://en.wikipedia.org/wiki/Tamamo-no-Mae
and is sometimes believed to be the same entity. I was into anime for a long time (Urusei Yatsura was a watershed but it really goes back to seeing Star Blazers on TV) but lately I have been into Chinese pop culture like Three Kingdoms, Nezha and Wolf Warrior 2 and that's gotten me reading about Chinese mythology and one clear thing is a lot of Japanese mythology comes from China, for instance it seems there is a "world tree" in almost every JRPG that they climb to get to heaven and even if they call it Yggdrasil it is not from Norse mythology it is from Chinese mythology. (Turns out also a lot of Chinese mythology as well as neopagan ideas comes from India as well.)
The Japanese do derive a lot of their culture from the Chinese, don't they? I would like to one day read some of the stuff around the Three Kingdoms. Alas! So much to read and do, and so little time!
Anyway, thanks for this conversation, I enjoy it.
The old adage applies: "everything nice is either illegal or bad for your health". Or both, I would add.
And is known to cause cancer in the State of California.
Never the middle ground, it's always a shocking new finding "by science" (spoiler: scientists seldom say the things newspapers and pop-science/nutrition & health articles claim they say).
I think peeing more lowers your uric acid.
Look in the medical literature and it seems outright spammed by reports on the positive effects of caffeine and negative reports on any of the harmful effects one would expect.
Similarly the main active ingredient of red wine (alcohol) is harmful, red wine in particular causes a lot of discomfort, dispepsia, hangovers and other unpleasant effects if you get a bad vintage but look the literature and it is like it will transport you to a blue zone and you will love forever.
And you find those kind of papers spammed in “real” journals, not MDPI or “Frontiers” journals.
Coffee prevents headaches for me so I'll always drink it. And no it's not related to physical dependence although at this point the withdrawal will guarantee a headache.
Got a mild flu/Covid/cold couple years ago. Better in a week. But, during the illness and since, the slightest bit of caffeine would make be incredibly wired to the point of panic attacks. Had to quit cold turkey. I’ve tried a cup now and again, and it’s the same thing: 6 hours of overwhelming anxiety.
Wierd. It’s like I became hypersensitive to caffeine. Oddly, though, nicotine doesn’t have that effect, and I always figured the two stimulants were similar.
I frequently switch between that and coffee (coffee has a much more pronounced effect and sometimes you have to grind)
Dose size matters.
I have the same issue w/ cannabis. Right now I have a few plants (legal) in the garden and also a bag that is going to a friend and I don't care. If I had a little puff though the next day I would want another little puff and another and in a week or so I would be like the guy in the Bob Marley song
https://genius.com/The-toyes-smoke-two-joints-lyrics
Oh, before I forget! Red meat and saturated fat is terrible for you! Wait, no, it's actually sugar that's evil. Vegetables are great for you! Oh wait, most vegetables contain oxalates and other defense, mechanisms and compounds that are actually bad for you overtime.
Red meat is a little bit bad for longevity. The majority of the reported effect is correlative.
> it's actually sugar that's evil.
Sugar is bad but mostly because it's easy to overeat, and obesity is all around terrible for health.
> Vegetables are great for you! Oh wait, most vegetables contain oxalates and other defense, mechanisms and compounds that are actually bad for you
Cooking removes most oxalates (tho vitamins too, to be fair). But the overall effect of oxalates is relatively minor, except in extreme cases.
---
Every food source has advantages and disadvantages.
Not being obese is 75% of the health battle.
Skepticism is reasonable.
The only way (imo) to stay on firm ground is to acknowledge that someone published a thing saying xyz, and maybe that you are x% convinced by it. Can't get too far out over your skis going that route.
Something being in a category, such as "a study", doesn't tell you much about a thing. If you read multiple studies on vaccine safety critically and reason about them and what experts are saying about them, IMO most functional human being are going to reach the same general conclusion about vaccine safety. If you do the same thing on studies about seed oils or aspartame you're also going to come to the conclusion that they're safe! If you're not reaching these same results it doesn't necessary mean you're the one who is malfunctioning but you should seriously consider it and try again to learn what you might not know.
Do you think well-funded RCTs (like those that support vaccine safety) are just as weak as any old observational study?
But my question to the person saying it's problematic to defend vaccine studies and attack food results is: isn't it possible that you feel the research procedures used in one are superior to those used in another?
For example: vaccine safety study looks at 200,000 people and randomly assigns them to use or not use the vaccine. Coffee/red wine study looks at 30 people and surveys them about how they felt last week after drinking coffee/red wine. Looking at these two, I think it's fair to put more trust in the vaccine study.