Comment by questinthrow | Hacker News Reader

Comment by questinthrow | Hacker News Reader

I am by no means a china fanboy but there is a replication crisis across all countries and all disciplines, one example here: https://news.ycombinator.com/item?id=39098780

This is a generalized problem because back in the "good old days" science was done by a select elite of passionate individuals (think physics between 1900-1950). Nowadays science is dominated by midwits that were forced to do "science" because of societal expectations that everyone needs to go to university and become "educated". The end result is an endless tsunami of grifters and people who cargo cult science by making a few pretty graphs and dropping some massaged data sets so that they can get their pretty baubles and titles. This is the full extent of modern scientific thought for 99% of people doing "science". Of course the passionate 1% are still there pushing the envelope further but more often than not they are inconvenienced by the modern academic system and all that it entails (think Grigori Perelman or others like him)

That's survivorship bias. Of course the results that have survived a century of scrutiny look better than the average high-profile research published today. But there were also plenty of unremarkable and outright false results back in the "good old days", as well as ideologically motivated garbage such as scientific racism. Or Lysenkoism, which killed millions.

Much agreed with this. There is no area of science nor social science that we don’t understand better today than we did 70 years ago.

For one thing, we understand statistical methods much better. We have a much better sense of what theories only sound good and what theories actually have support in data. Even in psychology, we understand the problems with non-replicating results much better.

The callback to the good old days when only the elite had the privilege of contributing to science rings very off to me.

Or the guy in 1904 who irradiated beef bullion and thought he created life [0]

0: https://www.nytimes.com/2021/03/24/books/review/lifes-edge-c...

Good point. But please consider that certain cultures may be inherently more transparent and accepting of dissent (see global indices that measure freedom of speech and degree of liberal democracy). Plus, different cultures have different attitudes towards honesty and commitment to truth (e.g. https://www.nature.com/articles/nature17160). In addition, the last decade saw the emergence of colossal national programs that explicitly seek to inflate output metrics (e.g. China, Saudi Arabia, etc.).

I’d trust a paper coming out of the Netherlands way more than I’d trust one from China. For the same reasons I’d trust a paper from a reputable publisher over predatory one.

Unfortunately, you're not wrong. Since a long time, I'm convinced universities need to scale down, and quite a lot too. Higher education, as far as it's needed, should be relegated to schools without a research function. Universities should be primarily research facilities, only training those students who can become good researchers.

That doesn't solve the publication treadmill, but it would expect it to be a step forward.

Hyper competition and perverse incentives is present in western academia also. The only difference is in the amount, not if it is present or not at all. I don’t get how a bunch of HN readers could fall into the trap of binary thinking.

Replication crisis != poorly done studies != fake research. Related, but not the same.

Research may fail to replicate for many reasons. Especially in areas like social science. Test subject (person) randomly selected from a population in say, the '50s, was different from similarly selected test subject today.

Bad incentives & outright fakery just add to that mix of possible causes.

You clearly don't understand statistics. If a study is done such that this can be an issue, the study is at best useless

Good old days like lying about sugar and tobacco? The unscientific racism in the US and Germany? Lusenkoism?

If any, from 1970 we got RMS, Linus Torvalds, GCC, Emacs, Gnuplot, Perl, Python, BLAS/Lapack and so boosting scientific research to unthinkable levels...

Also quite hilarious that your example exemplifies my point that a scientific discipline like computer science has its peak when its still in the domain of a few elite contributors. Which for computer science was at its very start, compare that to how it is today when when we invent ever more byzantine ways to create CRUD applications since the masses have been attracted to "computer science" to make money.

Your 'elite' supported bullshit science not so long ago.

Even Edison with DC vs AC. Pick either that, or... the lead industry. Or tobacco. Or...

More: https://news.ycombinator.com/item?id=39500484

Yup there has been junk science in the past due to bad incentives. Thats why its a very good idea to double down on these counterproductive incentives and also push everyone with a pulse into university so that he does cargo cult science based on these incentives because in this way we will uhh... make science better or something.

What I notice a lot happening is established researchers with reputable pasts wasting funds on research directions that are basically designed to fool the funding agencies and their bosses and their naive PhD students, but which to anyone initiated in the field are obvious dead ends.

The problem is not one of "midwits". The problem is that in science one is supposed to be in search of the truth and nothing but the truth, but we do not reward honesty. We only reward new discoveries or what appears to be new findings. Then it is no wonder that people are being dishonest.

Would second these observations …

Lol. The history of science is one of extreme graft. The idealized old days you present bear no relationship to what happened.

You might need to take another humanities major before understanding why I put the good old days in quotes, dont forget to publish a few papers on it to get that citation count up though!

Reply on news.ycombinator.com