No More Shall I Roam

In a recent experiment, subjects played a computer `game' where points were scored completely randomly, having nothing to do with what the players did. One group randomly got points 33% of the time, the other group 66% of the time. After 50 or 100 trials, the players were asked, amongst other things, for helpful tips for incoming players to score more points. Most of the players -- including even those who only got points 33% of the time! -- had helpful tips for incoming players, stuff that had `worked for them'.

I mention this because I find I'm spending more and more of my blogging time inveighing against anti-science, taking some inspiration from the effortless ease with which Canadian Cynic sticks it to evolution denialists; but I've never really gone into why science is important, and why it is probably the only way available to humans to understand the world around us.

How could the subjects in the above experiment have gotten it so wrong? Their actions had no impact on the results, yet they had crafted theories about how the point-scoring worked in the game, and believed that they had tested those theories. So what happened?

Humans are incredibly good at seeing patterns. Children as young as six months old can recognize faces, even faces of different species. But the downside is that we are so good at it that we find these patterns -- or faces, or `lucky streaks' -- even when none exist. The region of the brain which is responsible for seeing patterns can be imaged while the subjects watch lights flashed randomly, and when easily recognizable patterns that arise by chance -- repetition or alternation -- are broken, a strong reaction is seen. People just want to find patterns in events. This tendency to see patterns is called apophenia.

Now, seeing these patterns -- even if some aren't actually based in some significant underlying cause -- isn't necessarily a bad thing if you're (say) trying to figure out how to score points in a computer game. There are a variety of patterns presenting themselves to you, and you have the opportunity to see if they hold. Unfortunately, we humans have a long list of biases which shape our decision-making, and can sometimes mislead us. In the case of the computer game experiment, the most insidious is confirmation bias; the tendency to search for and be more convinced by evidence which supports a theory we're looking at than one that opposes it. When you're in the heat of a computer game and you find that going left left up down gets you a point, you may not really remember unless you're taking notes that you had gone left left up down just 10 moves ago and nothing had happened.

These two together -- seeing patterns everywhere, and being swayed by tidbits of confirmatory evidence -- are why the gamers thought they had insights into the scoring of the game to share.

"Science is what we have learned about how to keep from fooling ourselves." -- Richard Feynman

The old, dusty recitation of the scientific method that we were taught by rote-- theory, hypothesis, experiment, data, analysis, conclusion -- really is the key to how this process works, although it's a much more lively, combative, creative process then most of us got to see in school. A theoretician, trying to understand some observations, comes up with a theory; experimentalists (and now, computationalists) come along and say ``That's a lovely new theory you've got there; it'd be a shame if something happened to it. The most immediate consequences of your new theory are surprising results A, B, and C. Let's have a look, shall we?''. They go off and check to see if A, B, and C are true. There is no greater joy for an experimentalist than breaking a theory or finding an unpredicted phenomenon, and none for a theoretician than to see a theory take off and explain things beyond what they originally imagined. (There's no greater joy for a computationalist than to find someone who'll actually hire a computationalist, but that's a post for another day.)

This interplay is an essential part of doing science. Science was the original Open Source. Ideas and data are shared openly, increasing the ability both to build on others work and to tear it down. The Open Source software mantra, ``With Enough Eyes, All Bugs are Shallow'' in science reads something like ``with enough scientists, flawed theories are done for''. Although confirmation bias (say) may make it hard for one scientist to see problems in a beautiful new theory, another scientist in another country may not care for the theory at all and have no problems finding (and demonstrating) flaws in it. It doesn't matter how influential a figure you are; if your theory says A should happen, and people can demonstrate that A, in fact, doesn't happen, then your theory is broke. Maybe it's something that can be easily fixed, or the whole theory needs to be tossed, but you aren't going to fool anyone by claiming that the theory, as it is, is still good. The blogosphere should be so self-correcting.

Science is nothing more than a formalized, institutionalized way of critically testing new ideas. Ideas that work, pass; ideas that don't work eventually fail, although they may survive for a time. The fact that false ideas might survive for a while, even with rigorous testing, is disappointing, but the alternative to science -- the alternative to constantly putting ideas to rigorous tests -- is to allow all flawed theories to survive. People or administrations that dismiss science, or peddle pseudoscience -- and there are a lot of them, and they are well funded -- are hoping to make critical thinking unfashionable, and the rigorous testing of flawed ideas questionable. The only possible reason to do this is because they have flawed ideas that they hope to propagate.

I don't know if we will ever have a complete, true explanation for everything -- how the universe and the world work -- but I do know that unless you're willing to test, and test hard, every idea that comes down the pike and throw out the bad ones, you'll have lots and lots of falsehoods to choose from.

// posted by Jon Dursi @ 1:42 AM