Ideally, one adopts a hypothesis specifically to prove it false, not because one is sure it is true. Even if you believe it before you run an experiment on it, you do so provisionally, without putting "faith" in it, because you know you might be forced to discard it. This is not the faith in science you are looking for.
I've generally found that the expected mechanism comes first; an opposing hypothesis to falsify comes after. With complex scenarios it's can be necessary to develop in depth the consequential outcomes before one can find something that can readily be demonstrated empirically.
How do you determine otherwise which experiments are worth performing. We don't have infinite resources.
For example. Einstein published his theory of relativity in 1916 having worked, apparently, on it for 8 years. Do you suppose he worked on it because he believed it false? One of the outcomes of the theory was then first observationally confirmed by Eddington's observations of an eclipse in 1919.
Up until 1919 Einstein and other's put much faith in the theory by using it to model reality without knowing if it's predictions would be observed without falsification.
I'm all for Popperian falsification as the modus of formal scientific advance but on an individual level it appears to me that those who work for many years on unconfirmed theories are often believing in the truth of those theories without empirical confirmation.
>"Even if you believe it before you run an experiment on it, you do so provisionally, without putting "faith" in it, because you know you might be forced to discard it."
Moreover, it strikes me that one could not put faith in anything (the love of your mother say) if it were prevented by the possibility that later evidence would force you to discard it. That statement also appears contrary to the illustration I've given concerning GR too.
I think that is from some unrealistic picture of a scientist that probably comes from a standardized public school science curriculum. It seems to be an "ideal" that no one lives up to, and so when otherwise-uneducated people meet real scientists who obviously don't live up to it, it probably gives "scientists" a bad name to them. This irritates me, so I shall try to state my view, which I think is a better vision. (It more accurately describes real people; however, you said "ideally", so that objection doesn't apply; however, I think yours is, further, the wrong ideal.)
Usually, one adopts a hypothesis as an attempt to understand the world, which is usually an attempt either to satisfy one's curiosity or to better direct one's efforts. Then one tries to test the hypothesis to see whether it's true, hoping on the one hand that the world can be easily understood or that one's plans can be easily fulfilled, but on the other hand not wanting to be burned by investing intellectual or physical resources into false beliefs.
Attempting to falsify possible hypotheses is one method of trying to get at the truth, but by no means is it the only one, nor is it always the best one; e.g. when the main challenge is coming up with a good theory (e.g. relativity; heh, the sister comment gave the same example), you will not get there by trying to falsify hypotheses (it might start you out by showing that existing mechanics is wrong, but it won't help you come up with relativity). It seems that a focus on falsifying hypotheses would be mainly useful when you're given a lot of hypotheses from untrustworthy sources--or, at least, sources who aren't good at doing the kind of fact-checking that you can do.
I'll give an example from earlier this evening: I'm waiting for the new Plinkett review video to come up on redlettermedia.com (I expect it to appear sometime tonight), and instead of repeatedly manually checking the page it'll show up on (which I would otherwise do), I want to write a shell command to check it for me (currently checking every 30 seconds). I decide on this method: Use curl to get the text of the webpage, save it to a file, and compare it to a previous version using "diff"; if the files are different, then that hopefully means the new review has been posted.
However, I know that some websites will generate unique things for every visitor--maybe increment an "N people have visited this website" counter or display the current time or something. I don't want this script to give false positives, so I check. I use "diff -u" to compare two versions downloaded within a few seconds of each other. And the result:
Looks like it generates some unique token per visitor. Ok, I use "grep -v data_track_clickback" to remove that line, and now consecutive versions of the page are completely identical. I write up the loop and run it.
What did I do here? I hypothesized that consecutive versions of the page would be identical until the new review was posted; I hoped this was true, because then my shell command would be easy to write. However, I knew there was a chance it wasn't true, so I investigated and found that it wasn't; I then modified my hypothesis/filled in an alternative hypothesis (I had foreseen the possibility and what I would do about it), that "consecutive versions of this page piped through grep -v would be identical until the new review was posted". By design, this hypothesis passed the test that the previous one failed; I stopped investigation at this point and simply acted as though it were true.
This further hypothesis could have been wrong: as I wrote this comment, I realized it could have displayed the hour, or even the day, and then it would give a false positive but wouldn't do so until somewhat later. Also, it could be that my "grep -v" command would remove another line that would eventually contain the new post; I thought of this but deemed it unlikely and didn't bother investigating it. (So far tonight the script has correctly given negative results; this level of certainty was good enough for me.)
How does this fit with your narrative? I didn't adopt my first hypothesis specifically to prove it false. I did, at least, proceed quickly to test it and to prove it false. But then there was my second hypothesis. I definitely didn't adopt that one specifically to prove it false, nor did I even try to falsify it.
Perhaps you would say, well, you're not acting as a scientist. Perhaps not; I have a definite goal in mind, getting a script to work, and I'm interested in the truth only insofar as it helps me. Well, except that I'm also interested in practicing my skill at automating things, and in understanding webpages and other aspects of the real world so that I can automate them and do other things with them.
Fundamentally, we want to know things because we have some use in mind for the knowledge, or because figuring things out has been useful in the past and so it's associated in our minds with a positive feeling (some of which is probably evolutionarily hardwired into the brain), or because we want to exercise our skills (which, again, has a probably-somewhat-evolutionarily-hardwired positive association in our minds). I think that anything that could be called "pure" research, done without practical results in mind, falls into one of the last two categories. (I also think that even if your research's only purpose is to develop technology, when you are actually doing the research, your brain will work best when your overriding emotion is curiosity.) My point is, science is not divorced from the real world, but fundamentally dependent on it; and if your ideal of science is so divorced, then attempting to reach your ideal (which is the point of an ideal) will eventually bring you misery. Exactly, in fact, the way that a false belief will: holding X as the ideal of Y is believing that the best possible Y is X.
Scientific publications in different branches of science have different standards of evidence. Why? Because of the real world; because in some fields, experiments are expensive and difficult to conduct, and the main experimental tool in some fields is statistical studies that inevitably have some uncertainty. The standard for publication is, ideally, "Would the other scientists find this useful to read?", and when they can't do better than "The null hypothesis would cause this to happen with <5% probability" plus a plausible explanation plus documentation of what they did, that sort of thing passes the standard of publication. Does this make such investigations not science? (Some might say yes.) No. Science is the attempt to understand the world; any paper made with that intent that helps other scientists in their own attempts is a good scientific paper. (A paper that accidentally helps scientists could be called evidence or inspiration.)
