[sytelus]

I call these hard science vs soft science. Please do not get me wrong, I shuffle from math and art quite a bit these days and get to take a look at various approaches in both areas. There is a trend of applying hard science approach to soft science. This is good thing until it results in research that is quite wishy washy. When you see "study" where experiment is mainly surveying people, its soft science. Unfortunately such soft science gets a lot of press. When you see articles like "Coffee found to reduce cancer by researchers in a study", you are reading soft science. A huge problem is that people doing hard as well as soft science are referred to as "researchers" or "scientists" and both of their works referred to as "science". But its quite not. Science demands not only evidence but understanding. If coffee was found to reduce cancer, can you elaborate exact mechanism? If you can't then science requires not to make such claims. Doing survey of graduate students is not sufficient. That's the nature of hard science which soft science (including economics) violets too often. Media needs to be more aware otherwise the trust in science that has been developed over centuries will no longer be there.

[kissickas]

You had me for a second with the survey comment, but we don't know the mechanism for a lot of biology and that doesn't mean it's not a science. In fact, we didn't understand most of the things we take for granted today but that doesn't mean there were no scientists until Newton came along (or any other arbitrary point of "understanding").

[leeoniya]

i think the monimum bar for something to be science is that it has to have a rigor to it which can be used for removing doubt:

- hypothesis

- control group

- well chosen or random samples from representative population

- statistical significance

- a way to separate correlation from causation

- enumeration of conflating factors and potential for flaws in the chosen methodology

- list of prior studies or research

- peer review

- reproducibility, possibly using alternative methods

until all of this is done, a survey (or any study) is not science. convincing the layman is insufficient; you have to convince other experts in the field.

[sytelus]

In a way, science is a process of creating a model. To create a model, you first need evidence (i.e. data) and then you form hypothesis which your proposed model. Then you make prediction using your model that wasn't known before. If predictions continues to remain true over time then you have higher confidence in your model. However, a true scientist would never set his/her confidence to 1.0 in any model because all models are eventually wrong and needs to be improved further. So the science is the process of continuously gathering evidence, improving model and remain skeptical that you might be wrong. It is very much like training a machine learned model using training data. Most soft sciences do the first two steps and bypass everything else. It's like you created ML model, you had good result on training data but you never tested your model on hold out set, assumed your model was good enough and just moved on to make a press release.

[sytelus]

Biology had always been red herring. Lot of biologists actually are careful about making big claims until they understand the mechanisms. There is an evidence based clinical science but that's been made very watertight through structured trials that must comply to well defined standards unlike studies like "Coffee was found to reduce cancer" which is often half-assed surveys with too many statistical biases.

[kissickas]

We don't know the mechanisms of plenty of things that we accept as proven with scientific principles, like most of pharmacy.

And what about physics before Newton? Or before Einstein? In fact, we still don't know exactly how gravity works. Or magnetism. At what point are you drawing the line of being able to "elaborate the exact mechanism"?

[TheOtherHobbes]

You never elaborate the exact mechanism. But if you can consistently make accurate predictions, you're onto something.

The real argument is about the validity of the predictions. Core science - basically undergrad - is very good at making predictions in its domains of interest. Outside of that everything gets more speculative.

The real problem with soft science research is that it cargo cults data -> statistics into data -> weak correlations -> "truth." And that's not how good science works - because there's just a statement based on correlations that may be accidental, and there's no attempt to make a model at all.

[unishark]

The hard versus soft science distinction is an old one, cutting off around biology. Basically physics is the "true" science and as it gets harder to relate phenomenon back to the underlying physics involved, the study is seen as more soft.