SUMMING UP

CHAPTER 1 BEGAN WITH TWO QUOTATIONS about spec­ulation in science—one from Newton and one from Einstein. Newton, in his “official” methodology, tells scientists never to speculate.

My first concern was to define speculation in a way that could help to provide these answers. I proposed that speculation, in the “truth-relevant” sense, be construed as introducing assumptions, under what I called “theorizing” conditions, without knowing that there is evidence for those assumptions. The question then becomes: What concept of evidence should be used? Two probabilistic concepts, Bayesian (B-evidence) and my own (A-evidence), were introduced. My concept—there are, in fact, different objec­tive and subjective versions of it—employs the idea of the probability of an explanatory connection between evidence and hypothesis. The Bayesian (increase-in-probability) con­cept does not, even if the latter is changed to require that evi­dence increase the probability of the hypothesis to more than one-half. The explanatory idea, I claimed, enables us to see why certain assumptions can be regarded as made very prob­able by certain facts (e.g., facts about what the authorities believe), even though the assumptions are, and ought to be, regarded as speculations—that is, ones lacking evidence.

My approach to speculation in science is pragmatic. It rejects rigid views that say “never do it,” or “do it with abandon, but test,” or “do it with abandon, and don't worry about testing.” Whether to do it, when, and how depend on the context of inquiry. It depends on: (i) what you know, or at least think you know; (ii) what you are trying to explain, unify, calculate, predict, and so forth; and (iii) how you are proposing to do so. If you are Newton employing his three laws of motion plus theorems, attempting to explain the motions of the planets and their moons by establishing the force(s) producing these motions and the governing law(s), then speculation is not what you want to do. (Newton did it anyway, or so I claim.) If you are Maxwell attempting to see whether a mechanical explanation of known gas laws is even possible in terms of molecules in motion subject to forces, then do it by making assumptions about the motions of the molecules and forces acting on them, even if you have no ev­idence that molecules exist. Don't worry about testing, since (in 1860) you can't do it anyway. Your job is to theoretically develop such a theory, not to test it.

Similarly, how to evaluate a speculation is to a consid­erable extent pragmatic and context dependent. A specula­tion can be evaluated in an epistemic way, a non-epistemic way, or both.

To take his rightful place among the scientific immortals, Newton never needed that!

Armed with the conception of speculation I propose, and with a pragmatic attitude toward speculating, I examined some very general claims made by scientists and philosophers about the role of simplicity in science, about whether sci­entific statements can be tested individually or only “ho­listically,” and about whether a “'lheory of Everything” exists and why it would be good to have one. Despite the pronouncements made by supporters of these claims, the claims themselves are all speculations. They lack sufficient evidence to be believed. The question then becomes: Even though they are speculations, are they good ones? Is there any reason, whether evidential or non-evidential, to offer in favor of making them? Do they have to be made to do and understand science?

Both Newton and Einstein claimed that nature is simple, and therefore, that simplicity is an epistemic virtue. The simplicity of a theory, law, or explanation is a guide to its truth.

Theories that make fewer assumptions, postulate fewer dif­ferent types of entities, and have simpler equations than other more complex theories are easier to use for explana­tion, prediction, calculation, and communication. Most im­portant, as I illustrated with Maxwell's 1860 kinetic theory, they are excellent starting points for developing ideas that, although more complex, are likely to be more reflective of reality. The simplicity of Maxwell's early kinetic theory is not a sign of its truth, nor did Maxwell take it to be. He explic­itly regarded the theory as a speculation, and he introduced simple assumptions in order to enable him to develop them more easily and see whether a mechanical theory of gases was even possible.

The debate between “holists,” such as William Whewell, and “particularists,” such as J. S. Mill, contains various speculations about what counts as evidence for a scientific theory and about how one is supposed to evaluate a theory. These speculations, which I claim are unwarranted, are made by both holists and particularists, even if the latter are some­what less vulnerable than the former. In the mid-nineteenth century, Whewell said that evidence for the wave theory of light—evidence from phenomena such as diffraction and interference—was conclusive.

Whewell's idea is that for observational results O to constitute evidence for a theory H consisting of a set of assumptions, O must be derivable from H. Since such derivations usually require all the assumptions in H, it is the set H as a whole, rather than individual hypotheses in H, that is supported by the observed phenomena in O. Mill is saying that this is not enough for evidential support. Each assump­tion in H must be supported by some or all of the phenomena in O. Whewell's underlying evidential claim here—that in sci­ence, evidence involves only a deductive explanatory-predic­tive relationship—is a speculation. Worse, it is one subject to numerous counterexamples. The concept of evidence I have proposed (A-evidence, in its various forms) contains the idea that evidence involves some type of explanatory relationship between evidence and hypothesis.

268 | SPECULATION: WITHIN AND ABOUT SCIENCE a type of particularism and reject Whewellian holism. Is this enough to reject holism? No, because the holist can reply that scientific claims, including evidential ones, can be defended only holistically. This is yet another speculative assumption that I argue is without merit.

A major problem I find with the holism-particularism debate is that its focus is entirely epistemic. Even as a purely epistemic debate, its main focus, in the case of both Mill and Whewell, is on one evaluative question: Is there evidence suf­ficient to believe the theory? (For Mill, there is only if there is such evidence for each assumption. For Whewell, there is only if consilience and coherence have been satisfied.) To be sure, both writers allow the additional epistemic question: Is there information that makes it reasonable to treat the theory as a possibility worth considering? But these questions seem to exhaust their epistemic repertoire.

The alternative I propose is more pragmatic. It rejects the idea of focusing exclusively on epistemic evaluations. There are non-epistemic ones that, depending on the con­text, can be quite important. And even within the epistemic evaluations, the labels “proved” or “unproved” are often in­sufficient to describe the epistemic status of a theory. In ac­tual practice, as demonstrated by examples from Maxwell's evaluation of kinetic theory and Baden Powell's evaluation of the wave and particle theories of light, theories are often in a state in which neither of these labels applies. They are in a state somewhere in between “proved” and “unproved.” And even if one of these two labels can be used to describe some of the assumptions of a theory, there may be others, including the theory as a whole, that cannot be so described. What evaluative categories to use depends both on the status

of the theory and the context of evaluation. If you are writing a brief history of nineteenth-century physics, “unproved” may be the best term to use for wave and particle theories of light in 1833, and for Maxwell's kinetic theory in 1875. If you are writing a more extensive history, or if you are Powell in 1833 or Maxwell in 1875, a more informative fine-grained evaluation is called for.

Even if not all the assumptions of a theory have been “proved,” one can still use the theory, and scientists fre­quently do, to explain, predict, calculate, and so on, while admitting that perhaps some but not all the assumptions used to generate the explanations, predictions, calculations, and the like have been proved or established, or are better defended than others, or even have no support at all. The same can be said about the explanations, etc. generated from those assumptions.

The final speculation I consider is that there is a “Theory of Everything” (TOE) to be found and that it would be good for scientists to try to find it. Such a theory would explain everything by appeal to a set of fundamental entities (ones that have no constituent parts) and a set of fundamental laws (ones that cannot be explained by anything further). The claim or assumption that there is such a theory has been made by various scientists and philosophers, including some con­temporary string theorists, seventeenth-century mechanical philosophers, Thomas Nagel in defense of a panpsychism, and David Chalmers in his “construction of the world.” The claim or assumption is a speculation in my sense, since the supporters of this idea do not know that there is evidence to support the general idea, nor do they have evidence for any specific theory they propose as a TOE. Indeed, in the case

270 | SPECULATION: WITHIN AND ABOUT SCIENCE of Nagel's panpsychic theory and Chalmers' “construction of the world,” there isn't even a specific theory being proposed. Nevertheless, supporters of the general idea do present epi- stemic and non-epistemic reasons for believing that a TOE exists and that its existence must be assumed to make science possible. I considered historical, presupposition, unification, and a priori and empirical “strategy” arguments for the ex­istence of a TOE. None of them rises to the level of evidence (potential, ES- or explanatory B-) sufficient to believe that a TOE exists. Nor do they even provide reasons sufficient to take the idea seriously.

I also discussed the related normative idea of TOE supporters that it would be good for science if a TOE did exist, since then the world would be completely intelligible. There are two ideas of “complete intelligibility.” One invokes an assumption of a “rational order” for everything in the uni­verse, whether or not scientists can know what it is. The other contains the idea that the world is completely intelligible to us (or to scientists). I argue that it is this second sense that is of most interest to scientists. I also argue that with respect to this kind of intelligibility, what needs to be defended is an idea illustrated in the work of both Newton and Maxwell, which I call “Newtonian intelligibility.” An explanation can provide complete intelligibility for some group of phenomena without providing it for all phenomena. And whether it does so for the phenomena in question is most usefully judged by reference to contextually dependent standards of com­pleteness, not to some idealized, unrelativized one. Such a judgment takes into account the scientist's task: what he was trying to explain, how, and how well he has succeeded. In such a case in determining whether, or to what extent, the

scientist's explanation provides “complete intelligibility,” we ask whether or to what extent his task was completed successfully.

Newton clearly believed that he had completed his task successfully. From the motions of celestial bodies, using his three laws of motion, he thought he provided sufficient evi­dence for a universal force law that explains celestial and ter­restrial motion. He admitted that although he produced such evidence for the law, he could not explain why it holds. Yet, he said, what he accomplished was enough to complete his particular task in the Principia. Yes, there are future tasks to be completed, depending on the questions being asked and evidence that is, or will become, available. New evidence may suggest how the new task is to be accomplished, or indeed new evidence (not available to Newton) may show that the original task is based on faulty assumptions. We can under­stand Newton's claims of successful completion in terms of subjective evidence. If we employ an objective concept of evidence, such as veridical or ES-evidence, we have to con­clude that Newton's task was not complete. The evidence he supplied for the law was neither veridical (since the law is false) nor ES-evidence (since it wasn't sufficient).

Even if we employ one of the objective concepts of ev­idence and regard Newton's task as incomplete, and hence the “intelligibility” he provided for the phenomena he was attempting to explain, as incomplete or worse, the main point remains: neither Newton's task nor intelligibility, if any is produced, needs to be regarded as complete or completed only when he or others have constructed and provided ob­jective evidence for a TOE. Completeness does not have to be judged, if at all, on the basis of how close a theory comes to being able to correctly answer all questions that have been or might be raised, and doing so by appeal to fundamental “atoms” and laws. For “Newtonian intelligibility,” com­pleteness can and should be judged on a task-by-task basis without the completion of all tasks, and without necessarily completing the task in question by appeal to a TOE. Satis est.