I love this story so much I’m gonna tell it again:
I remember the Watergate thing happening when I was a kid, and I asked my dad, “So, when did you realize that Nixon was a crook?” My dad replied, “1946.” He wasn’t kidding. Nixon being an opportunistic liar was all out there from the very beginning of his career, and indeed this was much discussed in the press. Eventually just about everybody acknowledged it, but it took awhile.
When I posted this before, I listed a few examples where some people were able to stay afloat for years and years after their lying or fraud or misrepresentation or impossible promises were apparent:
– Theranos: the famed blood-testing company faked a test in 2006, causing one of its chief executives to leave, but it wasn’t until 2018 that the whole thing went down. They stayed afloat for over a decade after the fraud.
– Pizzagate guy from Cornell: people had noticed major problems in his work, but he managed to dodge all criticism for several years before being caught.
– Some obscure Canadian biologist: The problems were first flagged in 2010, this dude continued doing suspicious things for over a decade, and it finally came out in 2022.
There’s also that Los Angeles tunnel that didn’t make sense back in 2018 and still makes no sense.
And here’s another one:
Effective Altruist Leaders Were Repeatedly Warned About Sam Bankman-Fried Years Before FTX Collapsed
Leaders of the Effective Altruism movement were repeatedly warned beginning in 2018 that Sam Bankman-Fried was unethical, duplicitous, and negligent . . . They apparently dismissed those warnings, sources say, before taking tens of millions of dollars from Bankman-Fried’s charitable fund for effective altruist causes. . . . When Alameda and Bankman-Fried’s cryptocurrency exchange FTX imploded in late 2022, these same effective altruist (EA) leaders professed outrage and ignorance.
“Think long-term. Act now.” What could possibly go wrong??
I don’t have any deep theories about this one. It’s just interesting to me as another example where the problems were clear to people in the know, but because of some combination of personal/political interests and restricted information flow, nothing happened for years.
I think a large part reflects the sentiments of the “my wife thinks she is a chicken” joke (https://www.goodreads.com/quotes/208547-it-reminds-me-of-that-old-joke–you-know-a). This certainly fits the Sam Bankman-Fried case, but I suspect it fits the others to varying extents. Personal interest conflicts sound like the major factor to me. It is especially sad to see this in academic circles, where the stakes are often really low (promotion, tenure, status, etc.) compared with $billions.
There’s really not much filtering you need to do for someone donating money to you. If someone was lending money to SBF (as it turned out his customers were involuntarily doing), then there’s skin-in-the-game for him blowing up. Andrew’s view is that universities (and journals) lend their authority/prestige to researchers, so that stock can get degraded if lent out too freely to unreliable types. But I don’t think the same thing quite applies to recipients of SBF’s (now apparently ill-gotten) money.
> but because of some combination of personal/political interests and restricted information flow, nothing happened for years.
I think cognitive biases likely played a pretty big role. Which is just that much funnier since there’s linkage between the “effective altruism” folks and the “less wrong” “rationalist” folks for whom controlling for cognitive bias is a founding principle.
Oh no! More honesty researchers!
Well I guess it goes t’ show ya, when you believe in and promote BS, you need a steady flow of more and more BS to prop it all up.
I don’t understand your comment.
Me neither, but I think they might be making an argument that cognitive bias research might not replicate…?
Maybe “controlling” was the wrong word. I wasn’t referring to researchers, but a community of people who, as I understand it, rrecognize the impact of cognitive biases like motivated reasoning and try to follow certain practices to lessen their impact (like “steel-manning.”)
My point is that the cognitive-bias-claim industry is mostly cognitive-bias-driven to begin with. It goes something like this: “I don’t like what you [say, do, think…] therefore it not valid because it is driven by [your] cognitive bias”. Under those circumstances, it’s not surprising that people who view “controlling cognitive bias” as a founding principle would wind up having to be dishonest to support their views.
I thought pizzagate was not a Wasink thing (which I’m assuming is who you meant) but referred to the conspiracy theory about a child sex ring in the basement of a DC pizza parlor? https://en.wikipedia.org/wiki/Pizzagate_conspiracy_theory
Yeah, on this blog Andrew for a while has been referring to Wansink’s stuff as Pizzagate, it’s kind of an in-joke or at least jargon.
Ah, ok must have missed it. I was thinking that sure Wasink did some messed up stuff but he didn’t try to shoot up a pizza parlor as far as I know!
The costs of prosecution have an effect on the progress of justice (in many senses of the two key words). The kindest way I can think of putting it is that in a society with such a broad range of sensibilities, it must be practically certain one party’s bold claim will be another’s red flag.
The best lesson I learned from my day job is that every right is balanced against some other right.
For example, in some jurisdictions random drug testing on the job is disallowed because the individual’s right to privacy takes priority over the safety of the other workers, elsewhere it is the opposite.
Some academic institutions require scholars to sign a document saying they understand copyright law. Would signing such a document be an act of fraud in most cases?
The twist on the latter is that an academic maintained a list of predatory journals and publishers was threatened by one of the named publishers. I understood that the academic’s institution was not interested in defending the academic against the suit, the list disappeared.
The six month time delay in Andrew’s posting schedule means this post misses the Stanford president’s resignation.
See
The Research Scandal at Stanford Is More Common Than You Think
By Theo Baker
Mr. Baker is a rising sophomore at Stanford University. At its daily student newspaper, he won a George Polk Award for investigating allegations of manipulated experimental data in scientific papers published by the university’s president.
https://www.nytimes.com/2023/07/30/opinion/stanford-president-student-journalist.html
The scandal regarding the president of Stanford has to do with “blots.” Can someone explain in layman terms why “blots” are so important in science? I did find this
https://bitesizebio.com/10165/how-do-you-image-your-western-blots/
which says
“Developing your blot on x-ray film is the traditional way to detect protein signals” and is possibly enlightening to some people but not to me.
I also found this, which suggests why blot, whatever that is, is still around and in need of replacement:
https://www.science.org/content/blog-post/down-western-blot
“It’s for sure that characterization of proteins is a much harder business than the characterization of small molecules. I can see why the blot-image standard got started, and why it persists. But it’s too easy to fake, too easy to manipulate. Is there anything to replace it, or to enhance it to the point where you can’t just whip one up in ten minutes at the keyboard? I know that editorial staffs at the major journals have been thinking about this problem, but how much progress is being made? It’s still a key part of work at the bench, but when it comes time to publish, is it time to say goodbye to the plain old Western?”
In case anyone wants to delve more deeply into blots and Stanford, be sure to read about Elisabeth Bik
https://en.wikipedia.org/wiki/Elisabeth_Bik
who was at Stanford and
“is known for her work detecting photo manipulation in scientific publications,[2][3][4] and identifying over 4,000 potential cases of improper research conduct,[5] including 400 research papers published by authors in China from a research paper mill company. Bik is the founder of Microbiome Digest,[6] a blog with daily updates on microbiome research, and the Science Integrity Digest blog.[7][8]”
“Bik was awarded the 2021 John Maddox Prize for “outstanding work exposing widespread threats to research integrity in scientific papers”.
On Western Blots:
Very short: https://www.genome.gov/genetics-glossary/Western-Blot
Wikipedia’s article is ok, but I increasingly think that Wikipedia science articles are becoming bloated and poorly organized: https://en.wikipedia.org/wiki/Western_blot
There are different types of blots but this looks like it is about western blots, which measure protein.
Imagine the cell is like a house. The proteins would be all the stuff that makes up the frame of the house, the appliances, sinks, toilets, tables, trash cans, beds, and so on. Ie, the structural and functional elements.
The house equivalent of doing the blot is breaking down the house into the various components then sorting them by weight. Then you can further distinguish between parts with similar weights by checking some other property like if it is magnetic (for the blot that is done with antibodies that primarily stick to specific segments of proteins).
The importance to science is the blot lets us study the amount of various components present. This can differ due to the type of cell (eg, skin cell vs liver cell) and various conditions (eg, healthy vs diseased). The idea is if there is way more/less than usual of certain components that tells us about what the cell may be up to.
> Can someone explain in layman terms why “blots” are so important in science? I did find this
A blot is a way of determining how much of different types of protein there are in some sample.
First you extract the protein and place it in a gelatinous material, then you apply a constant electric field to the protein, which causes a current to flow. The proteins become charged and flow with the current through the thick gel. Smaller proteins flow faster… So you spread the proteins out in space based on their size.
Then you run the protein up or down out of the gel onto some kind of “blot paper”.
Then you add an antibody that binds to the protein of interest. Then you add a second antibody that binds to the first antibody and has some way of developing a color/stain or radioactive source.
So then on this 2D image of “lane” vs “size” you have also dark color. The presence of dark bands at a certain location along the “lane” indicates presence of a protein of a given size which is bound correctly by the antibody, indicating the presence of a specific protein.
Of course, all of this is just “a dark blob on an image” which is trivial to fake in photoshop. The forensic techniques for detecting this fake are not something normally taught to biologists, and certainly wouldn’t be something done routinely by PIs who run labs. As I’ve said before it’s just **too** easy to fake bio data so bio labs run on the assumption that people aren’t faking things. If they are, it’s possible to eventually detect, but imposing the necessary level of constant scrutiny would make bio labs a hostile workplace and no-one would do bio work anymore, so it’s not really a viable alternative.
Actually you make all the proteins have approximately the same (negative) charge by adding a detergent called SDS (you will probably see it in your shampoo ingredients as sodium lauryl sulfate). It unfolds the protein then binds and stabilizes the unfolded form. The SDS is bound in such large amounts the original charge of the protein becomes a rounding error. You can see a rough diagram here:
https://en.wikipedia.org/wiki/Sodium_dodecyl_sulfate#Laboratory_applications
They are separated by size because the smaller proteins can pass through both small and large pores in the gel, while larger ones can only migrate towards the positive electrode when they happen to get pulled into a large pore.
Imagine a thousand/million layer galton board with various bead diameters and spacing between the pegs. Then there are little motors in the beads so they vibrate enough to rarely get stuck for long. If you take a snapshot after some minutes the smaller beads will tend to be closer to the bottom than the larger ones.
General technique explained well, but it not just for proteins. The method is also widely used for nucleic acids as well. Actually the first methods, developed by Southern, used nucleotide probes to identify specific DNA fragments, and were labeled “Southern Blots”. Later the method was extended to stain for RNA fragments (“Northern Blots”) and lastly to identify proteins by antibody labeling (“Western Blots”). Initially the staining was an autoradiogram produced by the labeled probes containing radioactive isotopes that exposed a film to produce the spots which were then scanned. Now most systems use a chemiluminescent tag that is directly visualized by a scanner. In some labs I have worked in the PI was very proactive on directly viewing the images and in lab meetings would want to see the original gel. Prior to Photoshop issues, the other ethical issue that would arise would be having the gels cut and pasted, for example to remove a lane that did not work from an otherwise useful gel.
In chemistry there have also been issues, prior to Photoshop, with spectroscopic images being altered with white out to remove unwanted evidence of contaminants in a new synthetic result. This happened enough to result in a joke about unrepeatable results being due to “X’s reagent” (whiteout with X the name of the chemist you did not trust.)
Actually, that should be negative charge per mass. The bigger proteins have more of the negatively charged SDS bound and thus are more negatively charged. So force and inertia cancel out to give an approximately equal acceleration towards the cathode.
Same with the Galton board analogy, the beads with larger diameter also have more mass.
And the NYT’s article has some really nice graphics showing how the images match.
This one popped up on my news feed today,
https://retractionwatch.com/2023/07/31/exclusive-how-a-dean-went-about-about-correcting-the-scientific-record/
a case where, when it was brought to the PIs attention that a postdoc had been faking the results, the PI asked the journal to retract the papers.
There’s still the question of why the PI didn’t spot the faking earlier, but it does look like thy did the right thing when it was pointed out to them.
Interesting story – this is the 2nd Dean in the recent past who encountered fake data, presumably due to underling researchers, who subsequently resigned their administrative position and returned to their faculty positions. This makes me wonder about the relative standards applied to administrators and faculty. Does it say that faculty are excused for sloppy practice but administrators are not? If it does, should it be that way? (I guess tenure makes it hard to remove the faculty without proving they faked the data, but administrators are at-will employees so can be removed for appearance of bad behavior or just bad publicity)
A fair amount of my published work has been about a widely used but bogus method for estimating the value of habitat in a stream as a function of the rate of flow, called PHABSIM (physical habitat simulation system). People used this to determine how much water had to released from dams to provide for the fishes downstream and that kind of thing. The method was developed by federal agencies during the energy crisis of the 1970s, when there was a rash of hydropower projects for which the agencies had to recommend flow release schedules, so it met a need, and it involved computer modeling and a lot of jargon that took some effort to learn, so not that many people understood it. Instead of fighting the method, the hydro and water companies hired consultants who argued for setting options in the program that would result in lower flow releases, so the method became kind of a vehicle for negotiation. A fair number of people realized early on that the method was bogus, but it was too useful for others to listen.
On of the problems was simply sampling uncertainty, but there was no easy way to estimate confidence intervals for the model predictions. My contribution was to realize that you could use the bootstrap to do that (https://afspubs.onlinelibrary.wiley.com/doi/abs/10.1577/1548-8659(1996)125%3C0458:LISMCI%3E2.3.CO;2), and that the confidence intervals for typical studies were so wide that the results were pretty meaningless. Needless to say, this result was not accepted by the people who made a living off the method. Somebody even published a paper saying that I misused the bootstrap because I used re-sampling with replacement, which was unrealistic because in a real study nobody would use the same data more than once. I had a hard time getting a response published, because the editor didn’t know how to judge which of us was right (https://scholar.google.com/scholar?hl=en&as_sdt=0%2C5&q=williams+bootstrap+sampling+is+with+replacement&btnG=). I had one reviewer argue that it wasn’t fair for me to try to take a way a tool people used for their work without providing them with another to use instead. There are other fundamental problem with the method besides the sampling that I’ve dealt with in a book, but people are still using it.
Variations of this story repeat themselves so often that I shouldn’t be surprised, but when I hear it it’s always just “wow”.
Most of the times I’ve tried to convince people to do something it’s met with similar stuff. I contacted one of my friends who worked in CA dept of water resources and had some discussions for example of doing estimates of dam and levee stability to help them prioritize where to inspect for stability. Her response was essentially “100% of places we inspect need major remediation, there’s no point in trying to prioritize it’s all trash and the only thing that controls is how much political backlash there is to attempts to force people to upgrade”.
😢
I feel your pain.
And the shredder saga continues…..
https://www.npr.org/2023/07/27/1190568472/dan-ariely-francesca-gino-harvard-dishonesty-fabricated-data
I think these cases proves that our modern society provides individuals the opportunity to disprove their critics, showcasing our societal tolerance. (Nevertheless, these frauds should have a shorter shelf life.)
Although it may seem self-evident retrospectively, isn’t this a kind of survivorship bias?
From the worse serial presidential liar, Biden in 1987, when he quit the presidential race for getting caught contently lying about his past: “I’ve done some dumb things,” Mr. Biden conceded at a stop-the-bleeding news conference at the Capitol. “And I’ll do dumb things again.”
At least he wasn’t lying about that.
https://www.nytimes.com/2019/06/03/us/politics/biden-1988-presidential-campaign.html
Same is true of Madoff. There was loads of evidence he was a crook. I know people who pulled their money from Madoff years before he was officially exposed, and a number of organizations refused to take donations from him because he couldn’t adequately explain how he was making money. I am inclined to think that a lot of people were thinking, “He is a crook, but he’s my crook”. But perhaps I’m being ungenerous and it’s just wishful thinking.