Too many labs run like this

The institutional problems go deeper than this. Of the graduate students who had the courage to do the right thing, how many of them are going to find good jobs, even the excellent jobs their show of integrity deserves? Not many. They don't have actual work/publications to show, they'll be labelled potential troublemaker, backstabber, damaged goods - or they'll just omit it from their resume to avoid to be tainted by association.

When what they deserve is, on the contrary, to teach the level where they lead a lab.

I think a lot of people in software development are familiar with the pattern described, of a person with power chasing the appearance of success without much regard to what's actually going on. In both cases I think we have 1) systems that often reward appearance; 2) power structures where one bad actor can override/silence objections; and 3) institutions that are very uninterested in the truth of how things went wrong.

I was once in a situation in my grad school years TA-ing for an abusive and insane faculty member. Nobody did anything until I told them they could keep my stipend for the semester and I’d pick up freelance work to pay my way.

Then suddenly I was showered with apologies and told it would be better.

Nobody ever stands up to these people, but they fold the second anybody shows a spine.

As an ex-academic, my personal experience was that two out of three academic labs that I worked in were engaged in fraudulent activities (but at the time, I didn't think to carry hidden microphones and microcameras around like a spy to document what was going on). The third lab was very high quality, though. The key differences? The good lab had:

1) Rigorous lab notebook policies - everything had to be documented and signed off on, and there the PI conducted a weekly lab notebook review with each grad student (note this was a fairly small lab, for some large labs this might be difficult, but in that case postdocs could review grad students and the PI reviews the postdocs). This led to one grad student getting booted from the lab for cooking data at one point.

2) Weekly lab meetings with review of progress and open discussion of any issues, such as uncertainties in data quality and so on, with all members present. No secret squirrel meetings or people trying to figure out how to cook the results for publication or patents or what not.

The frauds had either no or badly flawed notebook policies (record data using a pencil so it could be erased or altered later), a highly secretive modus operandi, lots of whispering in corners, grad students being pressured to get 'the desired result', etc. This goes on all the time, and generally only high-profile bogus claims are the ones that get any public scrutiny.

As far as why this behavior is common in academia? Well, if the PIs think they know how a system works then they can just invent data to back their belief and hope that someone else diligently replicates it, so they can be first to publish and get the prizes, career advancement, patents etc. that come with that. A fair number of big-name scientists have probably gotten away with this - an indication is that they make a big splash in a field and then never do much else with it. (At the grad student level, they'll invent positive (publishable) results just to get their degree and move on).

Solution: academics with government funding should be forced to endure audits of their lab notebooks and data retention policies by an independent party on a regular basis.

Discussion a few weeks ago: https://news.ycombinator.com/item?id=39642135

You'd think these folks would learn from Elizabeth Holmes that eventually, if you just make things up, you do eventually get found out.

I guess not.

It's really starting to get hard to believe ANYTHING that comes out in a research paper this days.

7 years ago Veritasium published a video showing that most published research is wrong. News since then has shown that it's not getting better.

https://youtu.be/42QuXLucH3Q?si=gu1di9idNBE9mXu6

Fake it 'til you make it.