SCIENTIFIC SPECULATION
that provides a reason why e or h is true, not just a reason to believe that one or both are true. Newton's evidence is the sort scientists subscribing to the “mechanical philosophy” wanted because, they believed, it is based on a probable explanatory connection between e and h: probably the reason why the planets sweep out equal areas in equal times is that they are subject to a force governed by the law of gravity.
My imagined Newtonian's “evidence,” even if it gives a reason to believe Newton's law, does not satisfy the condition that there is a probable explanatory connection between e and h. (It is not probable that the reason why the planets sweep out equal areas in equal times is that Newton believes this is true, or conversely.) My Newtonian's reason is an authoritative one of the sort noted in section 8. Even if “only-game-in-town” and meta-inductive reasons, like authoritative ones, are or can be reasons to believe that a hypothesis is true, they are not reasons based on a probable explanatory connection between e and h.Now suppose that, given Newton's “Phenomena,” the probability of an explanatory connection between these and his law of gravity is not very high, and that Newton did not know that there are other phenomena that would make it high.[35] In accordance with my A-concept of potential evidence, if Newton used the law of gravity in theorizing (which indeed he did, especially in showing how to explain the tides and other phenomena), he would then be speculating, since the “Phenomena” he cited would not be evidence for the law, and he didn't know that there are other phenomena that would be. Suppose, further, that a follower of Newton uses the law in theorizing, believing that Newton has evidence that provides a good reason to believe the law, although this follower has no idea what this evidence is. His reason for believing the law to be true is authoritarian—Mr.
Newton believes it. Now we can see how an important difference between my objective concept of A-evidence and the objective (and upgraded) Bayesian one makes an important difference concerning what counts as a speculation. According to the Bayesian, the follower of Newton would have evidence that the law of gravity is true (“Mr. Newton believes it,” which let us suppose raises the probability of the law to a number greater than 1/2). So, on the upgraded Bayesian concept of evidence, the follower of Newton would not be speculating.[36]By contrast, according to my A-concept of evidence, the fact that Newton believes the law of gravity is not evidence that the law is true (because of the lack of a probable explanatory connection between the law's being true and Newton's believing it). So, under the present scenario, since Newton did not have A-evidence that the law is true, our imagined follower of Newton would not know that there is evidence that the law is true, since his belief that it is true is based solely on the false assumption that Newton has evidence for the law. For him, as well as for Newton, the law would be a speculation, despite their protestations to the contrary.
What remains from the list of “ways to obtain evidence” in section 8 are Newtonian inductivism, the Whewellian version of hypothetico-deductivism, and Lipton's “inference to the best explanation.” Unlike the other three, all of these accounts involve the idea of some explanatory connection between h and e—e.g., that the reason that e is true is that h is. Of the “explanatory connection” accounts, I find Lipton's the most problematic. Suppose you can show that a system of hypotheses H, if true, offers the “loveliest” explanation of phenomena reported in e. Have you shown that e constitutes either A- or (upgraded) B-evidence for H? And if so, have you shown that the evidence supplied is sufficiently strong so that H cannot be regarded as a speculation? Have you shown, e.g., that, given e, the probability of H, or the probability of an explanatory connection between H and e, is greater than êã? To do so, you have to show, at a minimum, that beauty tracks probable truth—that the more beautiful a theory is, the more probable it is, and that if its beauty surpasses some threshold, its probability is sufficient for belief. Lipton simply assumes that this is so, without giving any argument.[37] This does not mean that beauty is irrelevant in the assessment of a theory, only that, if it is a virtue, it is a non-epistemic one: it does not, and should not, affect the reasonableness of believing a theory.
Of the remaining two “explanatory connection” accounts, I prefer the Newtonian to the Whewellian one. However, I will not pursue this here.[38] For the sake of argument, I will suppose that both Newtonian and Whewellian evidence e can make it highly probable that there is an explanatory connection between h and e (thus satisfying my A-concept of potential evidence). If you have A-evidence for h, then h is not a speculation for you. What is missing in “only-game-in-town,” meta-inductive, and authoritative “evidence” e (even if such “evidence” could make a hypothesis (more) reasonable to believe), and what is present in Newtonian and Whewellian evidence, is the idea that e and h are (probably) explanatorily related. With this, we have what I call “evidence,” or perhaps, more accurately, “explanatory evidence.” Without it, we don't. Explanatory evidence is what Brougham was demanding of Young's wave theory of light, Duhem of Kelvin's molecular theory of the ether, and Weinberg of string theory. These critics are claiming that the theories in question are speculations because their proponents lack such evidence.
This explanatory idea is built into my concept of potential evidence, but it is not built into the Bayesian B-concept, even if the latter is upgraded to require that e raise the probability of h to a point greater than Ó2. So, we can incorporate this idea into a Bayesian concept by saying that when one seeks explanatory B-evidence for a hypothesis h, one seeks an e that increases the probability that there is an explanatory connection between h and e so that this probability is more than Ó2. ^at is,
e is explanatory B-evidence for h only ifp(E(h,e)/e) > p(E(h,e)), and p(E(h,e)/e) > 1/2.[39]
If probability here is construed objectively, we obtain an objective concept of evidence. Now we have closed the gap very considerably between my concept of (“potential”) A- evidence and the Bayesian concept. The only important difference between my concept and explanatory B-evidence (construed objectively) is that the latter requires that evidence increase the probability of an explanatory connection, the former does not.[40]
Finally, then, suppose that, in the course of a scientific investigation, P introduces an assumption h under previously noted “theorizing” conditions.
Then, we can say that:(Scientific Spec): h is a (truth-relevant) scientific speculation for P (with respect to the truth of h) if and only if P does not know that there is explanatory evidence that h (A-evidence, or explanatory B-evidence).[41]
Even if ‘Theta-inductive” “only-game-in-town,” and “authoritative” facts could raise the probability of a hypothesis and make it highly probable, these “facts” would not constitute explanatory evidence for the hypothesis. If these “facts” are the only ones known that make the hypothesis highly probable, and if it is not known that there are any others that do so and constitute A-evidence or explanatory B-evidence, then the hypothesis, if introduced in a scientific investigation and believed to be true, is a scientific speculation.
It is this concept of speculation—(Scientific Spec)—that I utilize in what follows. I will take it to be the concept in question when disputes occur about when, if at all, to introduce truth-relevant speculations. It is the concept I will be employing in section 12, when I address the issue of how to evaluate such speculations.
11.