WHAT COUNTS AS EVIDENCE?

I will call this B- (for Bayesian) evidence in what follows. Depending on what sort of Bayesian you are, probability here can be construed either subjectively or objectively. In the former case, we obtain a subjective concept of evidence; in the latter, an objective one.

Elsewhere, using numerous counterexamples, I have argued that definition (B), whether understood subjectively or objectively, provides neither a necessary nor a sufficient condition for evidence—as the latter concept is employed in the sciences.^[16] I replace this concept with several others, the most basic of which I call “potential evidence.” It defines ev­idence using a concept of (objective epistemic) probability and a concept of explanation.^[17] 'lhe idea is that for e to be potential evidence that h, there must be a high probability (at least greater than Ó2) that, given e, there is an explana­tory connection between h and e. There is an explanatory connection between h and e, which I shall write as E(h,e), if and only if either h correctly explains why e is true, or e correctly explains why h is true, or some hypothesis correctly explains why both h and e are true. In what follows, I will call this and the related concepts I introduce A- (Achinsteinian) evidence:

(A) Potential Evidence: Some fact e is potential evidence that h if and only if p(E(h,e)/e) > Th e is true; and e does not entail h.

The concept of probability involved (objective epi- stemic probability) measures the degree of reasonableness of believing a proposition. I claim that evidence, and hence reasonableness of belief, is a “threshold” concept with respect to probability; and that when the threshold has been passed, e provides a good reason to believe h—where the degree of reasonableness increases with the degree of probability p(E(h,e)/e).

I will briefly mention three other concepts of evidence that are defined in terms of “potential evidence.”

(A') Veridical Evidence: Some fact e is veridical evidence that h if and only if e is potential evidence that h, h and e are both true, and in fact there is an explanatory connection between h and e.

Using (A') we can define a concept of evidence that is relativized to the epistemic situation ES of some actual or potential agent:

(A") ES-Evidence: Some fact e is ES-evidence that h (relative to an epistemic situation ES) if e is true and anyone in ES is justi­fied in believing that e is veridical evidence that h.

An epistemic situation is a type of abstract situation in which one knows or believes that certain propositions are true, and one is not in a position to know that others are, even if such a situation does not in fact obtain for any person.^[18]

Finally, as a counterpart to the Bayesian concept that is subjective, I offer this:

(A'") Subjective Evidence: Some fact e is person P’s subjective evidence that h if and only if P believes that e is veridical evi­dence that h, and P’s reason for believing that h is true is that e is true.

An example of all four types of A-evidence is given in note 20.^[19] The first three are objective concepts, in the sense that whether e is evidence that h does not depend on whether an­yone in fact believes that e is evidence that h. In this sense, only the last, A''', is subjective.

I much prefer the definitions supplied by the (As) to that supplied by (B). But since there are so many adherents to the Bayesian definition, I will make reference to that as well as to the (As) in my discussion of scientific specula­tion. Let's see how far we can get with either type of def­inition. Later, I will show why, in order to offer a more complete definition of speculation, the basic Bayesian definition (B) needs to be upgraded in a way that makes it much closer to (A).

5.