New Scientist Does trusting your gut make you unscientific? 18:00 13 April 2012
Andrew Pontzen, astrophysics postdoc I've been popping up at a few science festivals recently, discussing the _evidence for dark matter_ (http://www.cheltenhamfestivals.com/find-events/science/s3-dark-matters) with particle physicist Tom Whyntie. After a session at Cambridge Science Festival, UK, a man in a duffle coat approached me to explain that Einstein was wrong - that the universe is not, in fact, expanding. (Actually, a static universe would have pleased Einstein greatly, but that's beside the point.) As evidence, he tried to give me his book on reinterpreting the redshift of galaxies. When I politely refused, he accused me of being "unscientific". The implication is that being "scientific" means completely digesting and testing every idea before deciding whether it's right or wrong. But sometimes we have to make fast decisions based on prejudice, or we'll never get anything done. Is that OK, or does it fundamentally undermine what we're trying to achieve? Let's consider another recent challenge to Einstein: faster-than-light neutrinos. Forget what _we now know_ (http://www.newscientist.com/article/mg21328544.700-neutrino-speed-errors-dash-exotic-physics-dreams.html) - even from the moment those results were announced, it was widely considered that they were unlikely to be right. But while the smartest people calculated the contradictions with results from _neutrino astronomy_ (http://neutrinoscience.blogspot.co.uk/2011/09/supernova-neutrinos-in-1983-and-1987.html) or the _disastrous implications_ (http://www.newscientist.com/article/dn21515-lights-speed-limit-is-safe-for-now.html) for basic particle physics, the more widespread reaction - _mine included_ (https://twitter.com/#!/apontzen/status/116975105950756864) - was simple scepticism. We can model this scepticism mathematically. The process of making rational judgements in the presence of uncertainty and incomplete knowledge is beautifully handled by Bayesian statistics. It's all about assigning numerical "confidence levels". If you assign 100 per cent confidence to true statements, and 0 per cent confidence to false statements, Bayesian statistics automatically boils down to normal logical reasoning. But in science we never have such clear situations; we have scales of plausibility. So, based on the reams of historical evidence, you might start by assigning 99 per cent confidence to the assertion that "Einstein was right and nothing goes faster than the speed of light". That means you are pretty confident it has to be true, but you allow a generous 1 per cent margin of doubt. Then someone claims to have measured neutrinos breaking the speed limit. How confident are you that the results from a typical, carefully conducted experiment are correct? Ninety per cent confident? That would mean that, typically, you take experimental results at face value. But this particular neutrino experiment has contradicted your 99-per-cent confident "speed limit" principle. Inserting all these previously settled-on numbers into Bayes's theorem, you find that it's roughly 10 times more likely that the neutrino experiment has gone awry than that the speed of light is genuinely broken. The confidence in a prejudice (despite being slightly eroded) has trumped the confidence in other people's experiments. Mathematics is hardly necessary in this case, but the point is that there is a watertight way to model the judgements taking place in our minds. We don't need to know anything about the details of the experiment. We can self-consistently use our existing knowledge to make a snap judgement about how likely the new result is. But does the existence of a coherent mathematical framework tell us that rejecting ideas - without even looking at them - is acceptable? Not really. It's a _useful tool_ (http://www.newscientist.com/article/mg20727725.700-cosmologys-not-broken-so-why-try-to-fix-it.html) as situations get more complicated, but it's still only modelling what we do in our minds. In particular, if we start with wrong prior beliefs, we'll end up making unreasonable snap judgements. For example, I could be blinded by my physics education, fooled into being far too confident about statements of orthodoxy. Once we admit that we really do operate in this world of personal belief, what makes us "scientific"? Essentially that we never assign any statement 100 per cent confidence (which would make it unassailable). We try to hold beliefs that can, given enough evidence, be overridden. For instance an additional, independent faster-than-light neutrino result would have beaten the 99 per cent Einstein-was-right assertion. We can then live with snap judgements and trust that other people will keep presenting us with new evidence if we really are wrong. I don't think I managed to convey all this very well to the man in the duffle coat in Cambridge. But then, he seemed to hold the unerodible prior belief that all scientists are lazily arrogant with 100 per cent confidence. So, by my definition, he's the unscientific one. Only if I'd said that, I might have appeared arrogant. -- Centroids: The Center of the Radical Centrist Community <[email protected]> Google Group: http://groups.google.com/group/RadicalCentrism Radical Centrism website and blog: http://RadicalCentrism.org
