Wednesday, January 1, 2014

Science and Reliability

Often enough, individuals and groups face decisions with grave consequences. An individual might be suffering from some challenging physical symptoms where drastic medical procedures are called for, and needs to decide which procedure is most likely to have the best outcome. The residents of a large region might be weighing the option of keeping a dam in placing or tearing it down. The decision might still be difficult even with perfect foreknowledge of the consequences of each option, but in ignorance of the consequences the decision is blind and the worst option becomes all too likely.

Typically each option will come with advocates who predict excellent outcomes for the option they espouse, while the advocates of other options will disagree with those predictions. Such disagreement is quite common. We need ways to evaluate predictions, to evaluate the arguments used to justify such predictions. We might want to compare the strengths of two opposing arguments. But if we can find ways to evaluate the strength of a single argument, then those can be used as grounds for comparisons.

In general an argument that action A will have consequence B should be based on prior experience with similar situations and similar actions. So the strength of the argument is based on the extent of such prior experience and how one builds the bridge from those past situations to the situation at hand.

Nowadays we have a marvelous systematic and institutionalized way of finding patterns that let us predict the consequences of actions: science. Science effectively addresses both these requirements of strong predictions. Science deliberately works to extend the range of experience, constantly exploring new realms. Science also constantly refines the theoretical connections that organize this mass of experience into a coherent framework. Science is a process that constantly improves the reliability of our knowledge of the world.

A particular prediction about the consequences of some action might well be characterized as scientific, but this does not imply that the prediction is particularly reliable, or any more reliable than some less scientific prediction.

Suppose, for example, that I would like to prepare dinner for a friend. Perhaps I have cooked a particular recipe for my friend on several prior occasions, with uniformly positive results. If I decide to cook this recipe again, it’s reasonable to expect similar positive results. Such a strategy could hardly be called scientific, given that it has probably been applied successfully for millennia, long before folks like Galileo and Newton who ushered in the Scientific Revolution.

On the other hand, suppose I find a scientific journal at the library and read a report that a common chemical constituent of many positively received meals is X. Supposing I have access to a bottle of X, I decide to reproduce the experimental procedures followed in the report and perhaps add chemical X to some plain biscuits and then serve these to my friend. I think it is not too unreasonable to claim that, however scientific this might seem, it is not a very reliable way to prepare a pleasing meal.

While science is, indeed, a process that improves the reliability of our knowledge of the world, for a prediction to be scientific is not the same as for it to be reliable. Even when a scientific prediction is the most reliable one we have, it still might not be very reliable. Because science is always extending the frontiers of our experience, there are regions in which we have very little experience. In those regions, scientific predictions may be the best ones we have, but they may still be very unreliable, i.e. the probability may be very high that the predictive framework will soon be revised as more experience is gathered.

One the other hand, most human experience is a natural evolution of the very complex interwoven network of human society, our local ecology, climate, etc. Doing science requires some level of repeatability, regularity, control, and precision. Much of our experience does not lend itself well at all to scientific analysis. Humans survived quite successfully for a hundred thousand years or more without any science to guide their actions, and other animals continue to live without science. Our ways of deciding how to act have evolved along with all the other facets of our being, and through that evolution have become well enough adapted to our environment that they are reasonably reliable much of the time. When we are acting in situations that are very much like those we and our forebearers have commonly experienced, our common sense pre-scientific predictions will generally be quite reliable.


  1. Is string theory science? Dark energy has an empirical reason for it to be theorized but string theory is glorified mathematical masturbation at this point.

    1. Yeah I am not trying to get too finnicky about exactly what makes for good science. A lot of high priced scientist do or have done string theory, so that's good enough for the point I am making.

  2. A nice & relevant article: