Some observations about evidentialism

February 2015

A killer argument against evidentialism is supposed to be: There is no evidence for it, and so by its own lights it is unjustified. But that depends on what we mean by evidence.

In legal contexts, evidence is defined by the law itself: Something is evidence if the law says it is, and otherwise it is not evidence. The law also declares certain kinds of evidence to be inadmissible, which is not the same as declaring it to be not evidence. In general, philosophers are not obliged to follow any legal dictates, but if we want our thinking to at least approximate that of ordinary people, we could do worse than take a few hints from what our courts of law are talking about when they talk about evidence.

If we do that, we get the notion that evidence is just any fact or set of facts that, in some sense and to some degree of plausibility, entail the truth of some proposition, e.g. "The person accused of this crime actually committed the crime." Lawyers typically distinguish between facts and testimony, but that is not a necessary distinction. Suppose a witness testifies that he saw the defendant commit the crime. That does not make it a fact that the defendant is guilty, but it remains a fact that the witness so testified, and this fact is supposed to make the defendant's guilt more plausible than it otherwise would be. That is why testimony counts as evidence. Nothing is to be presumed true just because somebody says it is, even if they say it under oath, but if somebody says X, that is one reason to believe X. Sometimes it is not a good enough reason, but it is always a reason.

It is for the jury to decide whether it is a good enough reason. To find a person guilty, they must decide that the evidence suffices to establish guilt beyond reasonable doubt. And that means all the evidence. Any particular piece of it might not be enough to preclude reasonable doubt, but that doesn't matter if all of it together is enough when weighed against any evidence for innocence that might have been presented by the defendant's lawyer. We are often reminded that no defendant has to prove his innocence, but that doesn't preclude anyone from offering evidence against their guilt if they have such evidence. This suggests two important points. A large number of facts sometimes can prove more than only some of those facts can prove, and there can be evidence both for and against some proposition.

That last point needs some emphasis. There are people who define evidence in such a way that there cannot be evidence for a false proposition: If some fact F actually is evidence for some propositon P, then by their definition of evidence, P must be true. Definitions are neither true nor false, but only more or less useful, and in my judgment this is not the most useful way to define evidence. It represents a kind of infallibilism that I regard as epistemically indefensible. It forces us, whenever we ask whether there is evidence for some proposition, to beg the question of whether the proposition is true. If the proposition in question happens not to be true, then any evidence that leads us to think it is true might be insufficient evidence, but insufficiency is not nonexistence.

As to what constitutes reasonable doubt, the law's position seems to be that that is for the jurors to decide. All of us have some notion of what reasonable people can and cannot believe, based ultimately, I strongly suspect, on what we ourselves believe and don't believe, and I don't see any good way around that. It does seem to force one of two conclusions, though. We must either suspect that everyone who disagrees with us about anything is being unreasonable, or else we must allow that on a great many significant issues, the most reasonable people in the world can legitimately disagree. Charity, obviously, demands the latter course. It's not that truth must be consistent with charity. It is rather that none of us can justifiably suppose himself to be so intellectually gifted that all other people should use us as the standard by which to judge themselves. I see nothing arrogant, though, about supposing that reasonable people should have some reason for believing whatever they believe. They can certainly disagree with us about whether their reasons are good enough to justify their belief, but to say they don't need any reason at all is to render the term "reasonable person" meaningless. At the very least, evidentialism says just that.

Most evidentialists, though, think it insufficient just to have some reason. Whether or not they've heard of Bayes Theorem, they embrace some version of it. Not all propositions are epistemically equal, and some need better evidence than others if they are to be justified. There are many things I can justifiably believe just because I hear someone say so. There are other things for which no person's testimony is sufficient evidence. This is a distinction we all make, even the most ardent anti-evidentialists. We do need to be careful with it, because it is susceptible to circularity, but the issue cannot be avoided.

The dictum "extraordinary claims require extraordinary evidence" is just a popularization of Bayes Theorem applied to propositions with a very low prior probability. Christian apologists typically sneer at it as if naturalists invented it just to discredit religion. Those apologists have no problem, though, applying it to any testimony offered in defense of any competing religion.

As with testimony, so with facts in general. They never speak for themselves. That is why a complete formulation of Bayes Theorem includes a factor for background knowledge. A proper evidentialist argument needs to demonstrate exactly why, all things considered, the evidence on offer makes the proposition in question more likely true than false, if that is what is being claimed, or else fails to do so, if that is what is being claimed.

But do we need a proper evidentialist argument for everything we believe? Why should we think so? Is there any evidence for evidentialism itself?

To my knowledge, no opponent of evidentialism claims we never need evidence or that, if we have evidence, we should ignore it. And there is a reason to accept evidentialism, which means there is evidence for evidentialism.

Assuming the contrary, there are some propositions we are justified in believing without evidence. What would they be? Some suggest that they include a priori truths or analytical statements such as mathematical facts or the principles of logic. I grant that such propositions are not supported by evidence as the word "evidence" is ordinarily used, but I don't agree that we believe them for no reason at all. It may be, and often is, asserted that we cannot doubt them. Very well. Our inability to doubt them is a fact. What better reason to do anything than that we cannot do otherwise? I don't think that is our only justification for believing analytical truths, but if we had no other, it would suffice.

Must something be true just because we cannot doubt it? No, but that is not the issue. The issue is whether we're justified in believing it, and whether that justification must consist of an appeal to evidence of some sort. As with evidence, we should not accept a definition of justification that makes it impossible for a false belief to be justified. Otherwise, any claim of justification is tantamount to a claim of infallibility.

An analytical truth cannot be false because to assert its negation is to assert a contradiction. This is not the case with synthetic propositions, and so they are never necessarily true, no matter how well justified they are. It could be untrue that the sun will rise tomorrow, but it is silly to suggest, for that reason alone, that we are unjustified in believing that it will rise tomorrow. Even Hume admitted as much, even as he despaired of finding a way to justify such inductive reasoning. Something that we cannot imagine might prevent the sun from rising tomorrow, but if it happens, it will not change the fact that we are justified today in believing that it will rise tomorrow.

What about moral truths? What evidence could I offer for saying that it's wrong to torture innocents? That depends on whether moral realism is true. If it is, then our minds are endowed with some kind of moral sense, strictly analogous to our other senses such as vision and hearing. My evidence for thinking something is morally wrong is just like my evidence for thinking that an object I see with my eyes is actually there: My perception is my evidence. And what if moral realism is false? In that case, my condemnation is just my expression of how I feel about it, and it is a fact that I have those feelings. I experience those feelings, and my experience is my evidence.

I believe similar arguments can be made for any other purported counterexamples involving propositions that I actually accept as true: They do not actually lack any evidence in the sense that we believe them for no reason at all. I don't have space here to address all instances that might be offered, and it would tax the reader's patience beyond endurance if I tried. And, I can always change my mind if someone shows me a counterexample that I have not already seen. At this point, I see no reason to accept the claim that there are propositions that we justifiably believe without evidence.

But I might seem to be begging a question. I claim that there is evidence for propositions that I believe, even though some people say there isn't any. Suppose we stipulate that those propositions do have evidence. What about propositions that I don't accept on grounds that there is no evidence for them? My adversaries claim that they are nonetheless justified in believing them. Do I have evidence supporting my belief that they are mistaken?

We are now in the epistemic territory of foundationalism. I claim that evidentialism is, in the usual lingo, a properly basic belief. In the terminology I prefer, it merits being regarded as an axiom that we should not believe anything for no reason at all. This is not the radical position that so many seem to think it is, given my preceding observation that we actually do have reasons to believe a large number of things that are commonly thought to have no evidence in their support. If someone believe in God because they cannot disbelieve, then their inability to disbelieve is a reason for them to believe. It's not enough of a reason for me to believe, but it is a reason to believe, and so it is evidence.

But now we're back to the problem of the usual meaning of evidence. In most contexts, when we're talking about evidence, we're talking about intersubjective facts, not personal states of mind. But also in most contexts, we seek to either justify or discredit some proposition that could in principle be intersubjectively obvious. We would not need trials if someone authorized to dispense justice were present whenever a crime was committed. Since that almost never happens, the prosecutor must convince a jury of what they would have seen if they had been present, i.e. that they would have seen the defendant committing the crime. The prosecutor does this by presenting some set of facts that make it unreasonable to think the defendant is innocent.

What makes his innocence unreasonable is a matter of probability. Sufficient evidence for a conviction is a set of facts that are collectively very unlikely to have obtained if anyone other than the defendant had committed the crime. The reasoning, if it's done right, is strictly Bayesian. The presumption of innocence sets a very low prior probability to the defendant's guilt. In order to derive a very high consequent probability, the prosecution must demonstrate that the probability of the evidence, given the defendant's guilt, is much higher than the probability of that evidence given the defendant's innocence, i.e. that P(E|H) >> P(E|~H). And it is the job of the defense attorney to show that this is not the case, that either P(E|H) is lower than the prosecution claims, or that P(E|~H) is higher. If he succeeds, then he has demonstrated that the prosecution has failed to eliminate reasonable doubt.

Of course the typical juror is not consciously doing a Bayesian analysis, any more than a baseball fielder is solving a differential equation of motion when he runs to catch a batted ball. But there is such an equation with one correct solution whenever a ball is hit, and the fielder must move according to that solution if he is to catch the ball. Likewise the facts presented at a trial will, in principle, justify only one verdict, and justice will be served only if the jurors reach that verdict.

There is no consensus on the probability corresponding to "beyond reasonable doubt," but it is generally understood to be not just anything above 0.5. Considering the cliche that we'd rather let ten guilty men go free than punish one innocent man, we should expect the threshold to be on the high side of 0.9. We can only hope that the average juror's intuitions about reasonable doubt are consistent with that standard.

In other contexts a lower standard can work. In a civil case, for instance, the "preponderance of evidence" standard in effect requires a Bayesian probability of more than 0.5: We should believe the plaintiff's claim if the evidence shows it to be more likely true than not. We thus accept a greater risk than we do in criminal cases that an innocent defendant will suffer unjustly because his suffering will not include loss of liberty.

This seems like a good general principle: We should believe a hypothesis if the evidence makes it more likely true than not, and otherwise we should not believe it, at least in ordinary situations. It's hard to see how a contrary principle could be justified. To believe something contrary to evidence is just to believe a contradiction.

This is not to say that all evidentialists, even if they're consciously Bayesians, will agree on everything for they have the same evidence. Bayes Theorem is not an algorithm of that sort. A Bayesian analysis requires judgments about various prior probabilities. Reasonable people can disagree about what those probabilities are and so will reasonably disagree about any consequent probability. But it does force us to make those judgments, and until we make them, we cannot logically infer any consequent probability. Without those judgments of prior probabilities, any claim we make that "The evidence shows X is probably true" is little more than guesswork. We are in no better epistemic position than the creationists who say that the Second Law of Thermodynamics is evidence against evolution.

(This page last updated on February 21, 2015.)

Site home.

Philosophy index.