HOW TO GET EVIDENCE

The answers proposed by scientists and philosophers over the ages have varied considerably and have sparked lively debates. I will begin with three historically influ­ential, contrasting accounts: Newtonian inductivism and two versions of hypothetico-deductivism—due to William Whewell and Peter Lipton—that introduce fairly strong conditions for obtaining evidence.

1. Newtonian inductivism. Briefly, following Newton's four rules of reasoning in the Principia, to get evidence for a causal law (such as Newton's universal law of gravity), you attempt to establish that a cause satisfying certain conditions exists (in Newton's case, that there is a gravitational force whose magnitude varies as the product of the masses of the bodies and inversely as the square of the distance between them), and that all objects of certain sorts (e.g., all bodies) satisfy a general law invoking that cause. For Newton, it suffices to establish that the cause exists by inferring that it does from the same observed effects it produces on various objects. And from these observed effects one infers, by in­duction, that all other objects of these sorts also satisfy the general law. This part of the argument Newton calls “anal­ysis.” The second part, “synthesis,” consists in taking the law and showing how it can explain and predict phenomena other than the ones initially used to provide a causal- inductive argument to the law.

2. Whewellian hypothetico-deductivism. William Whewell offers a sophisticated version, according to which to show that e is evidence for a system of hypotheses H, you show that H explains e; that H predicts and explains new phenomena of a different sort that are later established to be the case by ob­servation and experiment (“consilience”); and that the system of hypotheses is what he calls “coherent” and remains so over time as new phenomena are discovered. On this view, e together with the successfully predicted and explained new phenomena constitute evidence for H.^[27]

On both these influential accounts, if you want to obtain ev­idence for a hypothesis or system of hypotheses in the way each proposes, you must show more than simply that the hy­pothesis or system explains or entails some set of observed phenomena. For Whewell, you must show that it predicts and explains phenomena of different types, not just the sort you started with. For Newton, you need to do this as well, but also to provide a justified causal argument that the cause inferred exists and operates in a certain way, and a justified inductive argument that this can be extended to all bodies of certain sorts. Proponents of these accounts would claim that if you have satisfied the conditions they require, then the proba­bility of the hypothesis or system will be increased (indeed, to the point of certainty), thus satisfying the Bayesian defini­tion of evidence (B). Also, on these accounts, if you establish the hypothesis in the way proposed, you will have shown that there is a high probability of an explanatory connection be­tween the hypothesis and the evidence, thus satisfying my concept of potential evidence, given by definition (A).

The next account is similar to Whewell's in certain respects, but weaker.

3. Lipton’s “inference to the best explanation.”^[28] According to this account, to show that e is evidence for some hypothesis or system H, you show that H offers what Peter Lipton calls the “loveliest” explanation for e.

The last three accounts I will mention are weaker than those above. They are meant to provide only sufficient, but not necessary, conditions for obtaining evidence. And they are meant to provide sufficient conditions for obtaining at least some evidence for a hypothesis, not necessarily strong or conclusive evidence.

4. Meta- inductive evidence. The idea is that you can get ev­idence for a hypothesis if you can show that your hypothesis is of a certain general type and that other hypotheses of that gen­eral type have been successful in the past. For example, in de­fense of his molecular-kinetic theory of gases, Maxwell writes that it is a completely mechanical theory, and that (he claims) mechanical theories have worked well in astronomy and elec­tricity. These facts would constitute evidence for the theory.

5. “Only-game-in-town” evidence. You can get evidence for a hypothesis h if you can show that h is the “only game in “inference to the best explanation” is Gilbert Harman, in “The Inference to the Best Explanation,” Philosophical Review 74 (1965): 519-33.

30. See Lipton, Inference to the Best Explanation, 58; also, Achinstein, Evidence and Method, 95.

town” for explaining some set of phenomena e. It is the only hypothesis scientists have been able to come up with to ex­plain e, after thinking about the matter, and rejecting known alternatives for various good reasons. Then the fact that it is the only game in town is some evidence in favor of it.

6. Evidence from authority. You can get evidence for a hypothesis h by showing that the authorities or experts in the field relevant for h believe that h is true. The fact that the authorities believe h is evidence that h is true.^[29]

Defenders of the first three accounts—the Newtonian, the Whewellian, and the Liptonian ones—maintain that if you satisfy the conditions they advocate, then, given e, you will have shown that it is highly probable that there is an explan­atory connection between h and e, in which case you will have shown that e constitutes A-evidence that h. For the sake of argument, let us also suppose that, according to all six of the accounts, if you satisfy the conditions they specify, you will show that e increases the probability of h so that the latter becomes greater than % and e constitutes (upgraded) B-evidence for h.

9.