Apr 20 2009
An SGU listener by the name of Jenny recently sent me the following question:
When we skeptics are faced with an unexplained occurrence, we work up a list of possible explanations based on our knowledge of how the world works and then use Occam’s Razor to prune out the most unlikely ones.
We often preemptively wield the razor by leaving off the list supernatural explanations that have been tried and debunked dozens of times. This is a tempting time-saving measure, but it’s also a violation of our proclaimed skeptic’s creed of being open to ANY explanation that is supported by evidence.
I often wonder how Richard Saunders manages to stick to the creed with such good humor–I don’t have the patience, myself.
This misuse of Occam’s Razor is, in a way, the reciprocal of the argument from ignorance–the argument from presumed knowledge, let’s say. For we skeptics to state categorically that a supernatural explanation CANNOT be true is just as much a logical fallacy as for a credulous person to state that our naturalistic explanations cannot explain the phenomenon in question. It also leaves us open to accusations of arrogance and closed-mindedness, which in this case actually have some basis.
This is a great question. While Jenny is a skeptic who is just trying to understand skeptical philosophy, similar arguments are often used by the proponents of various supernatural explanations, and so a detailed answer is helpful on multiple fronts.
It is certainly a virtue to be open-minded, but trouble arises in how we define “open-minded”. It is often used by true-believers as an equivalent of faith, meaning that any odd belief must be accepted regardless of the logic and evidence against it. Whereas being open-minded in the scientific sense means treating all propositions fairly, without ideological bias. And then letting the empirical chips fall where they may – allowing science to function as a meritocracy of ideas.
In other words, we do not a-prior reject ideas, but once the evidence is in it is acceptable to reject failed notions in science.
Jenny is taking this principle but then making an unstated assumption that leads her astray. She is assuming that if we did not “preemptively wield the razor” by leaving off our initial list of possible hypotheses supernatural explanations, that the resulting list would be finite. Rather, if no filter or criteria were used in forming an initial list of hypothesis, the resulting list would be unending, limited only by our time and imagination.
One might then argue that only supernatural explanations which are already believed by some would need to go on the list, but this is just applying another criterion – popularity. It that not being closed-minded to ideas that are not popular? It is not the fault of an idea that no one has yet been clever enough to think of it.
We must apply some criteria in forming our list of possible explanations. Once this is recognized, we can then consider the appropriateness of popularity as a criterion. History has shown it is not a very good criterion.
What scientists generally do (and certainly within the applied science of medicine we take great pains to do) is to form a list of hypotheses from most likely to least likely. We then test the most likely explanations first (although sometimes we also go after the low-hanging fruit by testing the easiest to test hypotheses first, even if they are not the most likely).
But “least likely” trails off into infinity without objective end. So the real question is, how far down the list of prior probability are we going to go? That is a judgment call, but one that scientists have to make.
In practice what scientists often do is start testing hypotheses, starting with the most likely and most testable, until they find a hypothesis that is confirmed by evidence. But then they must also confirm this hypothesis by showing that alternative explanations are not true. But how many alternative explanations must be show to be untrue? As I argued above, the answer cannot be “all of them” because there is no limit to the number of alternate hypotheses unless we use some criteria of prior probability. In practice the answer is “all reasonable alternatives” with “reasonable” being a judgment call.
This is not only fair, it is necessary, otherwise science would grind to a halt testing an unlimited list of alternate hypotheses to each theory.
Further, it does not exclude even the most unlikely explanation from science. If such an alternate hypothesis turned out to be true, then the “more likely” hypotheses should all fail. Once scientists have exhaustively excluded their list of reasonable hypotheses, they will go back to the drawing board to see where they went wrong or too extend their list of hypotheses further down to previously considered “unlikely” alternatives.
Eventually they will get to the right answer, and they will already have done the necessary work of excluding more likely hypotheses.
The second problem with Jenny’s position is the claim that skeptics assume supernatural explanations cannot be true. Depending upon how one defines “supernatural” this statement may be little more than a tautology. But actually Jenny here is confusing philosophical naturalism with methodological naturalism.
The scientific method is dependent upon methodological naturalism – meaning that we cannot invoke the equivalent of “magic” as an explanation. This is because all ideas in science must be testable – there must be a way to falsify any scientific hypothesis with evidence. A supernatural hypothesis by definition cannot be tested because it is not contained within the laws of nature.
Therefore supernatural notions are not scientific hypotheses because they cannot be tested, and they therefore do not belong on a list of alternate hypotheses. These are the rules of science – if you don’t play by these rules, you are not doing science.
But “supernatural” does not merely mean currently unknown. Science explores news laws and new types of explanations all the time. Again, once we exhaust our list of possible explanations based upon current knowledge, we then need to seek new knowledge. This is often the most difficult, and creative, part of science – coming up with entirely new ideas and then (sometimes even harder) figuring out a way to test these ideas.
But science is actually agnostic toward the question of whether or not there are supernatural forces at work in the universe. Again – definitions get tricky here, because one could argue that any force at work in the universe is by definition natural. But let’s say that a supernatural notion would include the claim that there is an undetectable agency at work in the universe that could arbitrarily suspend the laws of nature. Such a claim is untestable. At best science could detect enduring anomalies – observations that forever defy scientific explanation. But science could never confirm that a supernatural explanation were correct.
From a practical point of view we keep coming back to the primary criterion of a scientific hypothesis – it must be testable and falsifiable. If it is, then it can go on the list of possible hypotheses. If it isn’t, then it is not scientific and it does not go on the list. The “natural vs supernatural” distinction is ultimately meaningless except for this feature. This is methodological naturalism, and science requires it to work. For any ideas outside of methodological naturalism, science does not hold that they are false, just unknowable to science.
Finally, it must be recognized that science builds upon itself. With each new question or idea we are not starting from scratch, as if we have no prior knowledge. It is not only practical, it is necessary, to approach questions in light of what has already been well-established. Only when that approach fails should we consider alternate explanations – but any testable hypothesis is ultimately fair game.
49 Responses to “Occam’s Razor and Closed-Mindedness”
Leave a Reply
You must be logged in to post a comment.