Why So Much Science Is Wrong | Hacker News Reader

Why So Much Science Is Wrong

aier.org

29 pointsRickJWagner13 comments

This smells like politically motivate BS. Science gets everything wrong, so it's OK if we just believe whatever we want to about the Pandemic.

For example, on the first page of this article, follow the link titled "We Must Choose: Liberty or Lockdown". There you will learn that the "CDC acknowledged that only six percent of US deaths came from Covid-19 alone."

You can find a good explanation of the facts behind this lie on Ars Technica. In short, on death certificates Doctors may list a Primary/Immediate Cause and any Contributing or Underlying Causes. For example the immediate cause of death may be heart failure with an underlying cause of Covid-19. Or a doctor could list Covid-19 as the immediate cause and something else as a contributing cause. In both cases, the person died of Covid-19. What the CDC reported is that on 6% of the death certificates doctors listed ONLY Covid-19.

But for 100% of those death certificates list Covid-19 as the cause of death.

What those assholes are saying is "Oh, your mother would be dead anyway - she had diabetes." When in fact she would NOT be dead except for Covid-19.

To quote Star Trek's Q: "Of course you'll still die. Just not right away."

Oh wow. Thanks for pointing this out. Even the first three links of the article are "What Greta Thunberg Forgets About Climate Change", "What Economists Can Teach Epidemiologists", and "An Avalanche of Failure" which is complaining about lockdowns again.

Then I looked up this "American Institute for Economic Research" and it's a conservative political non-profit dedicated to creating "a society based on property rights and open markets". That type of naming, trying to pass themselves off as a neutral research institute is some Prager U. level grossness.

As you said, the whole article is clearly politically motivated BS.

It would help a lot if more journals required preregistration of research designs.

In 2018 I heard an economist at UCSD who evangelizes preregistration list something like ten benefits. I can't remember them all, but among them are:

(1) It makes impossible most cherry-picking of one's data and model.

(2) Journals can choose to accept a paper based on the design, before the outcome, the eliminating publication bias toward positive or interesting results.

(3) Studies that don't get published in a journal are still recorded somewhere. (The preregistered design can be updated to include the results.) Not only does this prevent the loss of knowledge, but also unnecessary duplication of effort.

I haven't read this Wikipedia article on preregistration but I'm including it anyway just to be dangerous:

https://en.wikipedia.org/wiki/Preregistration_(science)

Much of bad “science” can be traced to innumerate psychologists typing their data into an R program, getting a p-value out, and then wildly misinterpreting that. They don’t know what a p-value is or what it means (not much, usually). What is a citizen, interested in keeping up with research results, to do? Remember what science is. Science is not searching for correlations in data. If you see a headline about a result that seems important, you can, you must, ignore it until you look at the research paper. It doesn’t matter if it’s not your specialty. Look for hypothesized mechanism and for successful prediction. If it’s not there, none of the reported associations or “statistical significance” mean anything.

Stats and correlations can still be useful. Quantum mechanics is a great example: we can use it to make the most accurate predictions in the history of the world yet most physicists would agree we have no clue as to the mechanism.

As I see it, the ecosystem around research journals is mostly to blame. Throughout history scientists recorded both surprising and unsurprising experimental results. Both were treated as important.

Now we basically discard and forget experiments / work that is not 'novel'. The bias introduced by doing this is what makes those p-values questionable. Furthermore, for some it seems that the goal has become publication rather than the search for truth.

Funny, there is a link in the article to a thing about eggs, and whether to stay away from them.

And it turns out it is all about cholesterol in them, as if that has anything to to with anything. The article starts put assuming dietary cholesterol is a real problem, and looks at whether eggs have any effect on that.

But current information is that eggs do seem to have something to do with heart trouble, just not via cholesterol. Now they think it is a consequence of what gut bacteria do to the lecithin and choline in the egg.

So they go on at great length about what is wrong with its treatment of cholesterol, all of it entirely disconnected from any reality. It is worse than what it criticizes.

Economics for you.

A better title would be "Most news articles on science are wrong, false...", and this article perpetuates that trend. I cannot fathom an expert in any field seeing one study and concluding "Well, I guess that's then end of that". They all know that getting the study published is a pretty low bar, and is only the beginning of the process, if that process is to be continued at all.

Yes, news stories latch on to the newest study all of the time, but that's not science, it's just one journalist sensationalizing one study. You will not see scientists in any field grab the newspaper to see what the latest findings in their fields are. They read the actual studies, and discuss with other experts in the field to determine if the findings are actually interesting or not.

> Going to the supermarket hungry makes you buy more calories.

> All wrong.

>misleading conclusions not warranted by the research itself; fabricated data; data manhandled to pass significance tests; incompetent experimental designs; or experiments that wouldn’t replicate when tried by other scientists.

Not wrong. But often experts proving what they think is true by the wrong means. The ear wagging the donkey.

I suspect this is how science in the past worked and worked well.

I think the explosion of scientists and the lower standards means this is now broken. But I don't think we can look to the scientific method to fix it, I'm not sure it ever worked as we think it did.

It's never the method's fault, it's always the implementation details.

Not quite. In the last 20 years, science work has been turned into scoreboarding: scientists accumulate points by publishing papers in scientific journals and getting cited, journals get points for being cited, institutions for keeping high-score scientists etc.

While the goal was clear (to establish a fairer promotion mechanism and remove personal bias), it has ultimately led to more people focusing on the score rather than the output. IOW, "gaming" the system. Even when it's done in good will, it might be optimising for the score instead of actual promotion of knowledge and science.

The way scientists are measured reminds me of how teachers get measured by their students test scores. In both cases we reward or penalise them based on some measure that is often only partially in their control, then people get angry when they try to maximise that measure.

The working scientists I know are as frustrated by the syst as it's external critics, but they still have to work in it if they want to pay their rent.

Reminds me of the saying, "as soon as a measure becomes a target, it ceases to be a useful measure".

> Not wrong. But often experts proving what they think is true by the wrong means.

Exactly, there is a big different between "this paper claims to show something that is actually not true", and "this paper claims to show something that it fails to show." They are both bad papers, but not the same kind of bad.

Reply on news.ycombinator.com