Evidence, Absence and Burdens
A lot of people think that in evaluating a given proposition p, there is a burden of proof such that in the absence of evidence, one side of the argument wins. This is commonly invoked when people say that positive claims have a burden of proof. What they are claiming is that if P is a positive claim, in the absence of evidence regarding P, it is appropriate to believe ~P. This is commonly brought up by atheists with regards to the existence of God. They argue that since we don’t think that Zeus, or Odin or Thor exist, we shouldn’t offer any credence to the proposition that Krishna or Jehovah exists either. But this argument presupposes that disbelieving in the existence of Thor, Odin and Zeus is the right option to take when we lack evidence one way or another about whether such beings exist. If we ought not to disbelieve in Zeus merely because we lack evidence*, disbelieving in Jehovah or Krishna because we lack evidence one way or another merely compounds the error.
The move to assign burdens of proof seems primarily suspicious because it involves apportioning belief*** (or degrees of belief) in a way that is not proportionate to the evidence while at the same time trying to lay claim to the mantle of epistemic rationality. But epistemic rationality just is believing propositions in proportion to the available evidence.
One rationale for thinking that positive claims have a default burden of proof is that it is impossible to prove a negative. Strictly speaking, this is not true. The violation of Bell’s inequality proves that there are no local hidden variables which determine quantum spin states. On a more mundane level, simply by observation I can prove that there is no horse in my bedroom to the same extent that I would be able to prove that there is a bed in my bedroom if I was seeing a bed in my bedroom****. The old saw that absence of evidence is not evidence of absence cuts both ways. Just as we should not treat the absence of evidence as evidence of absence, we should not mistake evidence of absence for a mere absence of evidence. It is not the case that we have no evidence regarding the existence of a teapot orbiting the earth. We could check with people in NASA for example. Did someone leave a tea pot in the air lock? It is in precisely the same way that there is strong evidence for the absence of a horse in my room. This is why it is appropriate to disbelieve a large number of mundane negative claims. Cases in which there is a genuine absence of evidence are comparatively rarer and somewhat more exotic.
Suppose it was true and there is an asymmetry between positive and negative claims. It doesn’t follow from the existence of an asymmetry that negative claims deserve some kind of handicap or head start over positive claims. Even if it was the case that some head start was required, the advantage negative claims have would at most be so slight that straight out disbelief in the face of no evidence is overkill.
Assigning burden of proof is not only restricted to questions of positive or negative claims, sometimes people say that the person who is trying to get the other person to change his mind has the burden of proof. This view is a version of epistemic conservatism. However, no account of epistemic conservatism is tenable.
There are three versions of conservatism, the weak, the moderate and the strong versions. On the strong version, for some proposition P having a belief P is sufficient reason to believe P when P is not the sort of proposition that could be analytically true. The moderate version states that having a belief P contributes justificatory weight such that the combination of reasons (P + R) is sufficient to believe P when either R or P alone is not. The weak version of conservatism is that some set of reasons R may be sufficient to maintain a belief P when P is already believed, but not sufficient to justify believing P if P is not initially believed.
The strong version of conservatism is absurd. Some R counts as a reason for P only if R is capable of providing occasion to revise confidence in P upwards. However, if R is identical to P, R can never provide occasion to revise confidence levels in P upwards. Thus, a proposition cannot be a reason for supporting that same proposition. The moderate version of conservatism faces the same problem, it is impossible for belief in (R+P) to increases the degree of confidence in P when R alone cannot. The weak version of conservatism faces a somewhat different problem. Consider the following situation.
John believes P on the basis of R such that if John did not possess R he would not believe P. However, if he did not initially believe P, having R would not be sufficient reason to believe P. John thinks weak conservatism is correct and also believes that R provides good reasons for his belief P. Consequently he must think that believing P is a good response to R. John believes that it is possible for R to be defeated even if he is currently unaware of any defeaters. John believes that if R is defeated, he will stop believing P. John also believes that if the defeaters for R are in turn defeated, he will obtain R again. However, John now believes that if this turns out to be the case, he lacks sufficient reason to believe P. Thus John also believes that believing P is not an appropriate response to the set of reasons R. John, therefore, believes that P is and is not an appropriate response to R. This is clearly incoherent. The only way to avoid such incoherence is to reject epistemic conservatism. If a given set of reasons R justifies a belief P, this set of reasons does so regardless of whether P is initially believed or not. By extension, two people who both possess R and have no defeaters for R should believe P. If all the reasons that they have are the same, then all the beliefs they have must also be the same.
At least as with regards to beliefs, assigning burdens of proof asymmetrically seems to be unwarranted. This should not surprise us. When you shift the burden of proof onto someone else, you are essentially claiming that you can rightfully believe some proposition even when there are no epistemic reasons supporting it. However, believing propositions for which there is no epistemic support is, by definition, antithetical to any reasonable conception of epistemic rationality.
One move that is often made in political philosophy that seems similar to the above but is not is the following: Exercises of coercion have a burden of justification which non-coercive arrangements lack. This sort of move is different because there is an antecedent principle such that exercises of coercion are prima facie wrong. I will leave the full argument for this to another post, but it would proceed roughly in the following way:
A necessary feature for a set of principle to count as the principles of justice is that they must realistically be able to effectively govern a well-ordered society. This requires widespread acceptance of those principles. However, being coerced, especially in ways that prevent one from pursuing one’s final ends, is deeply unpleasant and likely to cause a person to reject the institutions and principles that sanction such coercion. Therefore, unless coercion can be justified to everyone in society, it should be avoided.
Of course, if the above account ultimately proves faulty, then there would be nothing about coercion that required further justification.
Can any similar justification be provided for burdens of proof? Are there any other occasions which I have not covered in which the burden of proof allegedly falls asymmetrically?
*I am setting aside the problem of evil. If the problem of evil argument is successful, then there is some evidence against the existence of at least some versions of God**.
**I am not here to litigate the issue of whether God exists. I’m concerned with whether there is such a thing as a burden of proof that lies more heavily on one side than another.
***I want to be clear that when I talk about belief and disbelief, I am not only talking about the cases where my degree of confidence is 100% and 0% respectively. Of course if my degree of confidence is 100% I believe the proposition, 50% I am agnostic about it and 0%, I disbelieve it. As a very rough guide, anything from 0-33% is disbelief, 34-67% is agnosticism, and 68-100% is belief. This is rather crude, but good enough for our purposes now. You are invited to draw more distinctions if you so wish.
****caveats about evil demons, brains in vats and hallucinations apply where appropriate and symmetrically to both claims.