someone has dedicated a book (that unfortunately remains too superficial, however) to paradoxes and their importance in the advance of science

looking for paradoxes starts by questioning and experimenting

Replicability

what about paradoxes?

I wouldn't be too sure about that: a lot of so called social 'sciences' would rather classify as what Feynman called 'cargo cult science'. They do stuff which will look almost similar to real science, and they will also use 'scientific sounding' names like 'data analysis' for their work, but it often turns out that it is only a very bad emulation. Example: low sample number, no error discussion, no re-checking in psychology/sociology.



This is how science should be, but this is not the case in the real world, unfortunately. After all you must win grants and thus produce many papers, the content of the papers is not at all important for this. Also, many very specialized fields of science have a very small group of experts and the complexity of the field makes it effectively impossible for outsiders to verify their results. This of course leads to corruption. Also the peer review system is a joke. You pay big money to get your paper published, but the reviewer(s) gets 0$ - it is just a random contributer choosen by the journal - this again leads to corruption: either he knows the author (and then accepts if he likes him or not), or simply accepts for some high enough number of citations of his own works, etc - and I am not even talking about some small random crap journal, this is also true for high-impact journals like PRL. There are even journals/systems which actively promote 'wrong' science (which has been disproved by the experiment, like Bohm-QM, E8 TOE by that surfer dude) and are like a second, parallel scientific world with their own grants, journals ('the new scientist'), communities, which are hard to discern by an outsider, but also compete for real government grants, wasting the tax-payer's hard earned money.

what makes the evidence convincing? and how strong does it have to be for it to be convincing
for example if an experiment shown 80% accuracy would that be enough?
to whom do you have to convince?
Also i would like to add that different areas of science differ on convincing.

A lot of scientific work generally uses the 95% confidence interval. With that, you're basically saying that you're saying that you're 95% confident that whatever you're trying to measure falls within two values (that are mathematically calculated based on your actual measurements, so they're not just made up randomly).

You have to convince other people who are doing work in your field. Like for example, if you're a nuclear physicist and you say you have discovered cold fusion, you have to convince other nuclear physicists that your process truly works as you say it does. This is why it's called PEER REVIEW. Because your peers (people who do the same type of work as you) are the ones double-checking your work. It wouldn't make much sense for an astrophysicist to double-check the work of an evolutionary biologist, after all.

deleted

This post was edited by Nanako_Chan on Dec 7 2011 06:05am

The ONLY thing that makes science convincing is that it is testable.

And that makes it VERY convincing.

Maybe I wasn't clear enough. If you think that someone who is working in the same field as you is able to double check your work in a reasonably short time (please remember, you will get exactly 0$ for your review), you are clearly wrong. Modern science is so complicated that doing meaningful work requires so much specialization, that even your peers will not be able to fully understand it. Sure, they will probably get the general idea and can detect if something is totally absurd, but they will not be able to check every calculation without investing one week or so. And you can be sure that these little details are really important, and thus the peer review system is effectively worthless.
If you think this is not true, please goole for 'Bogdanov Affair'. And this is only the tip of the iceberg, everyone working in science who is honest enough can confirm that this is the preferred modus operandi for a vast majority of so called 'scientist' (especially in theoretical works without direct technical applications, obviously).

No matter the methodology you are employing there will always be variables and room for error. Someone could 'test' a hypothesis in a 'controlled' environment with unforeseen factors pre-existing and resulting from the experiment which conclusively alter the findings. Those alterations, variables, etc may never be fully observed thus science is far from perfect. If you carry out an experiment under precise guidelines with minimal deviation from the objective path 100 times and your findings vary only 1 time out of those experiments then it can be logically asserted your findings are not an absolute truth. Hence, science is not perfect and any ideology can be contested no matter how seemingly thorough or fool-proof.