What Science Can Be Trusted

One of the things I collect is stories of science gone wrong. Respectable, and still respected, scientific experiments that are deeply flawed and/or outright faked. I'm not alone in this since Richard Feynman taught himself not how to read bubble chamber photographs but how other scientists systematically misread them.

My suspicion of scientists started very early on when my high school physics teacher told me personally about how some students at the University of Toronto tried to reproduce Millikan's oil drop experiment with modern equipment ... and couldn't. In fact, not only was that experiment faked because the "results" were cribbed from theoretical values, but the theoretical value Millikan copied from was WRONG. As if that weren't bad enough, later scientists copied his "results" even when their own were more accurate. After all, it's not like such a renowned and well-respected researcher would have been a bald-faced filthy liar, could it? That's why the "empirically measured" charge of the electron shows a steady progression from Millikan's value to the true value over time.

The other story of shenanigans among scientists that marked me very early on was this story of a biologist who tried to make a rat maze experiment. So far so good, right? I mean, there are thousands of the fucking things. Except that he was obsessed with doing it properly. He wanted to eliminate every possible source of error and confusion. After a dozen iterations, he ended up with this kind of super-maze that had all kinds of insulating soundproofing anti-vibration features. That's great right? WRONG. Because what he did was invalidate years, decades, of other people's research. And he didn't even get any results from it. All he did was establish how rat maze experiments should be run. Wait wait, the best part's to come because you see he never got published. Yeah that's right, you can do first-rate science that invalidates thousands of other peoples' work and it isn't publishable.

More recently there was this fairly widespread story of how lab rats were being made sick by being fed standard rat food. Cause the rat food was made from soybeans. And if you know anything about nutrition, and aren't a braindead hippie, then you know that filthy estrogen-filled shit's horrible for you. These guys were testing cancer drugs if I recall correctly. While I'm on it, do you know why drugs that cure cancer in lab rats don't do jack in humans? It's because lab rats are really, REALLY prone to cancer. Animals that aren't hopelessly inbred and thus have functioning immune systems generally don't get cancer and don't NEED the anti-cancer drugs that work on lab rats. Well as if this weren't bad enough, it turns out the rat food had something to do with giving rats cancer too. So this "promising" anti-cancer drug turned out to do jack once the rats were given actually healthy food. The best part is that the filthy soybean shit they were feeding the rats was the same shit everyone else was feeding their rats.

Then there's medical experiments in humans. Those are a fun a dozen. Let's take breast cancer. The earlier you treat breast cancer, the better chances you have of surviving. It proves that early detection and intervention works, don't it? Not so! Cause there's this oft-forgotten thing called spontaneous remission. That's where your own fucking body naturally fights cancer all by itself and beats it. Many of the women who are diagnosed as having early stage cancer would have beaten it anyways. Without any treatment at all!! But forget that, let's just spin it as painful OUCH diagnostics and $$$ expensive $$$ treatments working! There's money in it, who cares about the truth? Kinda like the oncologists PRIDE themselves on planning anti-cancer therapies so that a patient gains, statistically speaking, a mere few days of extra life. We all know that a couple extra days of life are worth tens of thousands of dollars in the pockets of oncologists as well as excruciating pain for patients, right?

But there's no experiments like psychology experiments. There's the executive monkey experiment where two monkeys get zapped based on the performance of one monkey. The results of the experiment showed that the executive monkey got more ulcers. This is good, right? I mean it proves that managers DESERVE their ski vacations and massage treatments for deciding other people's fates. And we all love the rich, right? Only problem is with this whole "performance" thing. Apparently the researchers decided to choose monkeys for the executive slot based on intelligence. I mean, you wouldn't want a dumb monkey there, they'd get zapped all the time and it would make the experiment run longer! Yeah, so apparently after that little confounding factor got taken out of the equation, it turns out that, surprise surprise, the helpless monkey's the one with the ulcers!

Then there are experiments on hairless monkeys. Everyone knows of the Stanford Prison Experiment, right? Same with Milgram's Obedience Experiment. You know, the two experiments where you draft volunteers who are willing to obey the orders of some anonymous researcher and then you make them do horrific stuff, and then you conclude that ALL PEOPLE, regardless of whether or not they volunteered for psych experiments, are slaves to authority and would commit atrocities! Un-fucking-believable. You can't make this shit up. For fuck's sake, the experiments wouldn't have been conclusive even if they'd DRAFTED psychology students into them. Why? Because psychology students are abnormal (highly empathetic and irrational, generally incapable of logic) so they are not statistically representative of the general population.

Then there's the Six Degrees of Separation experiment. You know, the one where this bozo sent thousands of letters to be hand-delivered to a destination. Letters, 99% of which never got to any destination, but let's ignore that and focus SOLELY on the successes and then draw conclusions about the planet from it! Never mind that it became immediately obvious that people were stratified by class and that letters whose origin and destination were separated by class would just never get there. Or that hey most of the letters never reached their destination. Yes, let's make positive conclusions from utter failures! Unbelievable.

You know, there are monkey experiments that are fairly trustworthy. There's the Chicken Wire Mother Monkey experiment which determined that comfort is more important than food for infants. Funny how comfort isn't listed as one of the "16 basic needs" of humans, even though it's been known for centuries at least that human infants deprived from touch DIE.

Then there's the experiment where a bunch of monkeys in a cage were conditioned to beat each other up based on some signal, then they were rotated until none of the original monkeys in the cage were left. But every time the signal was given, the monkeys still beat each other up.

There's a couple things that make these experiments trustworthy. The first is that you're not pre-selecting monkeys. You have a bunch of monkeys and you just do something to absolutely every one of them. The second thing is that you're not watching for anything complicated or subtle. You don't care whether the monkeys play the violin or even whether they push a button on time, only whether they eat or they beat each other up. The last thing is that you're not depending on the monkeys to use their huge brains to learn and do something complex, you're looking at strictly animal behaviour. Simple experiments testing for simple behaviour are pretty reliable. Complicated experiments and/or complex behaviour are unreliable, no matter how spectacular they appear to be.

So what science can be trusted? Can you really trust those huge over-complicated equations in superstring theory? Actually yes, because math is simple. Math looks complicated to your puny, puny brain, but it's actually hella simple to mathematicians because it's regular and predictable. Well what about those huge experiments with those enormous overgrown particle accelerators at CERN and Fermilab? Surely that's too complicated! Surprisingly not since conceptually those are just hollow tubes drawn into a circular shape with magnets spaced a precise distance apart. The engineering might be complicated but the design is extremely simple. And there are thousands of engineers on those projects making sure that every single detail works to spec. Best of all, there are also thousands of scientists on those projects checking every little detail of the theory, including each others' work.

You see, "complicated" doesn't mean expensive. On the contrary. An experiment with a hollow tube in the shape of a perfect circle that happens to be 10 kilometers in radius is SIMPLE. The fact that it's expensive just means there's gonna be thousands of scientists to oversee this incredibly simple experiment. That's great! And going the other direction, a cheap experiment with a single human being, or even a fucking rat, is incredibly complicated. Because biology is complicated, because brains, even animal brains, are fucking complicated. And usually those experiments only have a single quack overseeing them. So expensive & simple == good. While cheap & complicated == bad. Which when you really think about it is terribly obvious, but people aren't used to thinking that a rat is complicated so anything at all you do with a rat is a horribly complex experiment.

See also Most Great Science Is Fraudulent... and Modern Scholasticism.