UNDERDETERMINATION
Underdetermination Claim: Scientific theories are underdetermined by the evidence. In order to choose a theory you must do so on the basis of simplicity: choose the simplest theory compatible with the evidence.
'1 his quite popular idea is that if we just look at the results of experiments and observations, these will be compatible with various conflicting theories that we have or might construct. So if we are to choose a theory, we must do so using some criterion in addition to empirical evidence. Simplicity is that criterion.
'1 here are two ways tounderstandthe Underdetermination Claim, depending on the purpose of the “choice” in question. If we are choosing a theory for non-epistemic reasons— e.g., to explain its historical importance or its fundamental ideas—then simplicity is usually irrelevant. If we understand the “choice” as an epistemic one (choose to believe, or just believe), then the question becomes: If a theory is underdetermined by the evidence, why should we believe the theory (i.e., believe it to be true or empirically adequate) if it is the simplest one compatible with the evidence? In such a situation, why is simplicity relevant for belief? The only possible answer is that simplicity is an epistemic virtue, that it provides an important epistemic basis for believing that the theory is true or empirically adequate (whether or not this idea is based on the claim that “nature is simple”). But this is just the epistemic (or ontological) claim that I challenged in the previous chapter, so I will not pursue it further.
Perhaps, then, the Underdetermination Claim should be understood as making a pragmatic assertion to the effect that since theories are in fact underdetermined by the evidence, and since scientists do need to choose some theory to work on, test, promote, utilize to make calculations, etc., they should choose the simplest theory for such purposes, even if simplicity is not a sign of truth.
I will challenge this claim by questioning the assumption it is based on, viz. that theories are underdetermined by the evidence. What exactly does this assumption mean?Here are two possibilities: (i) Empirical evidence by itself can never establish a theory with certainty, although it can make its probability high. (ii) No matter what the available empirical evidence, it is never sufficient by itself to make a theory more probable than not, or more probable than every competitor. Claim (i) is easier to deal with. If the empirical evidence can make the probability of a theory high (or, using my concept of A-evidence, if, given the evidence, there can be a high probability of an explanatory connection between the evidence and the theory), then even if this probability is not 1, it can be reasonable to believe the theory. Evidence need not be conclusive to make a hypothesis reasonable to believe. The fact that I own 95% of the tickets in a fair lottery is a very good reason to believe I will win, but not a conclusive reason.
What argument, then, can be given for claim (ii)? To take a simple example, suppose that a coin, drawn randomly from a coin machine in my grocery store, has been tossed in a fair manner 1,000 times and has landed heads roughly 50% of the time. Shouldn't this count as strong evidence that the coin is physically unbiased and that it will continue to land heads about 50% of the time if tossed? Shouldn't it make it at least more probable than not that this will be case, and hence more probable than any competitor?
You might reply, “No, there are various alternative hypotheses that are also compatible with the available evidence—e.g., that the coin is physically biased toward heads, not unbiased, but it is under the control of the Mafia, which makes it land heads approximately 50% of the time just during the first 1,000 tosses; or that the coin is physically biased toward heads in such a way that, in the long run, it will land heads 90% of the time, and the experimental result we got is just a matter of chance (after all, the probability of getting such a result with a 90% biased coin, though tiny, is not zero).”
Evidence about the results of the first 1,000 tosses does not rule out these conflicting hypotheses.
If we choose the “physically unbiased” hypothesis over the “biased” ones, we cannot be doing so on the basis of this evidence. There must be some other basis for this choice—e.g., simplicity. Simplicity is the sort of non-evidential criterion that must be appealed to in selecting what hypothesis to choose to work on, promote, and so forth. Evidence is never sufficient for this purpose.My response: “Yes, such competing hypotheses are ‘consistent' with the data so far, in the logical sense of consistency: there is no logical contradiction in conjoining the data with either of the two ‘bias' hypotheses. But there is no evidence that either of these alternative hypotheses is true—none that would give these alternatives any significant probability. And there is considerable evidence that makes it very probable that the ‘no physical bias' hypothesis is true: look at how and where the coin was selected, the way it was tossed, the way coins are normally manufactured, the fact that while the Mafia may be in involved in casinos, it generally doesn't get involved with coin machines in grocery stores, and, of course, the results of the tosses.”
This sort of reply is given by Newton in his Rule 4, in response to Descartes' attempt to crush or at least weaken inductive arguments generally by invoking alternative possible explanations of the data.[94] [95] If you are going to crush or weaken my evidence-based claim that the coin is physically unbiased by presenting other possibilities, you need to introduce evidence for these alternatives. Merely concocting some alternative explanation is not providing such evidence. The “alternative explanation” argument works only when there is either no evidence in favor of a preferred hypothesis or the evidence is pretty weak, or when there is at least equally good evidence in favor of the alternative. 2.