My problem with science

Yeah, yeah, I know, kind of like hating your own feet. But there it is.

I wrote this up on a message board group that I belong to. It's far from perfect - I neglected, for example, to address the fallacy of science-therefore-atheism, and the bit about science-as-faith is horribly undeveloped, and I was totally inadequate on the competing-hypotheses problem, and I really wish I had done the fallacious-reasoning-from-science problem more justice, but it's a good start.



Let me try to be more specific about what I think the problem is, in some kind of articulate manner. Please forgive me if I accidentally gloss over anything I reasoned through and then forgot which actually needs argument.

I'd better start with some knowledge I already know you have, in order to build on it. The irreducible part of science, below all the mythifications and sloppinesses and careerist trappings, is a method of extracting inferences about the world around us. Specifically:

1. Observe phenomenon (ie, a distinct and describable state or event - I should really get into phenomenology here but I won't)
2. Make conjectures about phenomenon
3. Identify circumstance under which, if your conjecture is true you'll get one result, and if your conjecture is not true you'll get a different result
4. Perform an experiment which:
a. Controls all other variables, so that differential outcomes can't really be arguably ascribed to outside causes
b. Varies the circumstance which is supposed to produce the differential result
5. Observe the results; if you get your finding, your conjecture *can be assumed to be* correct (unless a better explanation comes along), and if you didn't your conjecture needs to be modified or junked
6. Validate the results by replicating the experiment

Many conjectures together will produce a large body of inference which can in turn be subjected to further inferences. Those inferences will suggest circumstances, phenomena and conjectures which will together suggest a larger umbrella conjecture that makes some assertions about what's "really going on." Eventually enough small conjectures will be tested to (hopefully) show your larger umbrella conjecture to be robust and successful in knitting the small conjectures together. As part of the method, we can write this as follows:

7. Observe phenomena which you can relate deductively to previous conjectures
8. Make conjectures based on phenomenon + umbrella-conjecture
9. Test them per 3-6 above

Steps 1-6 are about hypothesis formation and testing, and steps 7-9 are about theory testing and formation.

***End stating of obvious***

Sorry about that. I have to spell out more stuff you know now. Again, sorry.

Karl Popper (based on Comte) did a brilliant job of showing that 1-6 are not about absolute certainty at all. They're about more-or-less-comfortably-establishing-reasonable-burden-of-proof. As per Aristotle and his statement "there are no black swans," it's essentially impossible to observe all phenomenological occurrences (and per Comte there's often no way of conclusively stating whether or not something that counts as a phenomenon in your knowledge class). Once you have, there's no way of knowing if *your* particular hypothesis is the only explanation; like the swans, you'd have to generate every possible hypothesis and test them all.

Thomas Kuhn did a similarly brilliant job of showing why steps 7-9 can be pretty limpy. Knowledge in motion tends to accrue to previous knowledge, due to human conservatism and desire to be right. Also when your head is filled with a thousand hypotheses it's pretty hard to see around them to the phenomena that might not fit and really need to be explained.

***Now, on to the argument***

Science as a pure method is one thing, but science in the world is a lot more problematic. We already have identified two problems:

1. Hypotheses are not the last word - there's no real way of knowing you have the right one, and whether or not the evidence for it is conclusive.
2. Theories are similarly flawed - it's likelier we have an incomplete one than a complete one.

To this, we can add some basic human fallacies:

3. When you have a hammer, everything looks like a nail.
4. Because of the strict controls that keep it robust and empricially-inductively valid, you have to apply science with vigorous constraints. No human is perfect at vigorous constraints.

Because of these two things, science is always going to be applied towards problems it can't possibly solve. Science will never be able to explain why men are assholes or whether there's a god. It may be able to generate a host of plausible theories as to why the first (unfortunately, without identifying the deductive fallacies inherent in the "logical" side of the argument, and without the necessary benefit of exclusivity), but it won't be able to provide an explanation. It won't be able to generate any useful information towards the second at all, because there are no observable phenomena that can only be explained by a God hypothesis - faith is per hoc unprovable.

People who are not scientists, and people who are, are going to get those three necessaries wrong over and over again. You probably know more examples of science gone wrong, and science gone silly, than I do.

Because science is a capitalist career dependent on making stunning discoveries, there are also these problems:

5. Peer review is not entirely objective.
6. Scientists will fudge to get their result, because they love their hypothesis or their job too much. (Look at David Baltimore for the most egregious of these, but it happens in many ways large and small.)
7. Most scientific fields will be hopelessly behind on step 6 of the method (reproducing results). [This last is not so bad, because there's implied validation in testing new hypotheses that work towards the same umbrella theory.]

Because science works on the absolute edge of what we know and what we think we know, there are also these:
8. Our measurement abilities lag far behind what we are trying to measure ("observe").
9. In keeping with problem #6 above, what counts as a predicted outcome is kind of a moving target. How much improvement do you have to see in a disease state before you can say that your empirical circumstance actually affects the disease? (This is especially true now that scientists are so dependent on statistics, which are only confident to a 95% level.)

Even with these nine problems, science is pretty damn good to us. It's real and it's effective, as Bryan so trenchantly noted. "It is better to be approximately right than exactly wrong," as John Tukey noted.

However, these problems speak to science as a *conditional but robust knowledge state*, which most non-scientists never seem to be able to wrap their heads around, and which leads us to the Big Dilly:

10a. Emprical hearsay is not deductive certainty. People who are not working in a particular field with which a theory concerns itself tend to ascribe a false state of certainty to scientific findings - there's no distinction made between the truth of the sun rising every day that we can remember, and the sun rising tomorrow; between global warming and a nice day out (or a cold rainy day, if you're in the anti-global-warming camp); between the time your girlfriend didn't get pregnant and the time she did. Worse, there's no distinction made between the theory of evolution and the theory of population dynamics, or the theory of Alzheimer's disease, or the theory of altruistic genes. Specifically, people are credulous about science - even scientists. Scientists working on AD are pretty dumb about global warming.

10b. Because so many mistake empirical hearsay for deductive certainty, there's a sense in which science is really an article of faith - viz, that we know as fact everything that is really critical, that we're learning more all the time, that we have the best explanation or are hot on the trail of it.

This, in my experience, leads to a knee-jerk smug superiority among many atheists, who assume they can explain the world better with the Universal God Science than the fundies can. Only they're not self-aware enough to call it God. And I don't think they're any more right than the fundies. Just more reasoned and more open to revision. To assume that the first has moral certitude is circular, since reason got us in this pickle in the first place; and to assume that the second has moral certitude assumes that people will give up on problems 6 and 10b.