h2g2 The Hitchhiker's Guide to the Galaxy: Earth Edition

In the scientific process, falsifiability is key to... to... it matters. So what is more important to science - proving something or disproving (its opposite)?

Y.

Disproving is more important.

Interesting question - made me think of my research at university, where I first proved that a particular n-dimensional quantity (n > 2) could be expressed as a certain nifty formula, and then I proved that the nifty formula was actually completely useless for n > 2...

well I think falsifiabiilty must be key, because you only need to find one example where it doesn't work to stomp all over a general rule

or something

Well, yes, falsifiability is important. Thus your claims have to be of such a nature that they can intrinsically be shown to be wrong. Compare "There are no white ravens" to "There are white ravens". Claim (1) could be falsified by finding a single white raven. Claim (2) cannot be falsified because even after a life time of seeing only black ravens, you cannot rule out the possibility that there are white ones and you just were unlucky enough never to find them.

But claim (2) COULD be proven to be true by finding a white raven. The means of disproving claim (1) and of proving claim (2) are identical. Just one white raven would do. So WHY is disproving more valuable than proving?

On a side note, why do they say that you cannot disprove a negative?

It might not be impossible to disprove a negative, but it seems very, very hard, which is why so many scientists seem to find themselves blindsided by things they hadn't anticipated. Like most of the rest of us, in fact.

I can prove that it doesn't matter

Can you disprove that it does?

Yes

Proving the null set [that the inverse or opposite of your theory is incorrect] can be a useful tool, depending on the type of thing being investigated, but there may still be limits.

Conclusive proof that [choose one: cholesterol, sugar, cigarette smoke, masturbation, etc.] is bad for you would only be possible in a universe where trial participants could experience a sequence of time twice -- once with the substance in question, and once without. Anything less than that begs the question of whether the trial and group and the control group are, for all intents and purposes, identical. You can, of course, get closer to the truth the more trial participants you have [40,000? 60,000? 100,000? The whole population of a medium-sized country?], but paradoxically the closer you come to getting a reliable truth from large numbers, the farther you get from applicability to the individual who wants usable data.

Cholesterol: a really large sample will usually show a small but real sample with very high cholesterol but who live well into their nineties in fine health.

Sugar: George Bernard Shaw, who had a wicked sweet tooth and lived a very long life. Unfortunately, you are probably not George Bernard Shaw.

Cigarette smoke: The two people who had the world record for longevity both smoked to some extent. I recall the Jeanne Calment gave up smoking when she was about 113 [I don't remember the exact age] because she didn't want to harm her health. She did get about 9m more years,. so it helped her.

Masturbation: well, the warnings about that were based on nonsense, anyway.

Who said that?



Deb

"Proof" in this sense is only applicable to mathematics, not any of the physical sciences.

You can't "prove" Newton's gravitational theory. All you can do is make repeated observations that it is correct to the level of accuracy of your observation. But that doesn't "prove" it. It just verifies that, for the observation you made, the model works.

All the physical sciences are like that - their "laws" are models of varying degrees of accuracy, and all it takes is one observation that doesn't fit the model to make it obvious you need a new one... which will just be a better model.

It's an interesting philosophical question - is there even any such thing, in principle, as a perfect model of the universe? As in, an equation or set thereof which will predict perfectly, to arbitrary accuracy, any behaviour of any system? Hope not.

Absolutely, what Hoo said.

Proof doesn't really exist in science, just more and more evidence to support a postulate.

Note that 'cholesterol' is NOT bad for you - it's an essential part of your metabolic system! http://en.wikipedia.org/wiki/Cholesterol

Too much of one type of cholesterol (low density lipoprotein -LDL) is the thing that gets the bad press.

"Note that 'cholesterol' is NOT bad for you - it's an essential part of your metabolic system" [Orcus]

Agreed. If my memory isn't too faulty, I believe that I've read about children with excessively low cholesterol and *extremely* short lives. Children with extremely high cholesterol seemed to survive longer. Maybe there are ways to filter out excessive cholesterol, but not add more when there isn't enough? For evolutionary purposes, things that kill young organisms before they reach reproductive ages are less likely to be passed on for the next generation.