Evidence and truth

Abstract

Among other interesting proposals, Juan Comesaña’s Being Rational and Being Right makes a challenging case that one’s evidence can include falsehoods. I explore some ways in which we might have to rethink the roles that evidence can play in inquiry if we accept this claim. It turns out that Comesaña’s position lends itself to the conclusion that while false evidence is possible and not even terribly uncommon, I can be rationally sure that I don’t currently have any and perhaps also that I won’t get any, and (absent certain evidence to the contrary) you are not afflicted with any either. This conclusion might seem too good to be true. I finish by raising a puzzle about one of the main motivations for Comesaña’s view.

Finding things to comment on in Juan Comesaña’s new book is an embarrassment of riches. There are too many interesting and challenging arguments to wrestle with. Here I will focus on a few points concerning a central theme, the theory of evidence. Perhaps the most surprising claim of the book is that my evidence can include false propositions. I’ll explore the implications this might have for the roles that evidence plays. Then I’ll go back and take a critical look at the main motivation for it.

1 False evidence

I’m looking at a wall that appears to be red. Nothing funny is going on as far as I can tell. I have no defeater for the belief that the wall is red such as an indication that there are red lights shining on it. According to Comesaña, my experience provides me with ex ante justification to believe that the wall is red. Furthermore, this justification is basic or non-inferential in that it is not based on my justification for anything else like that the wall appears red. On Comesaña’s account—he calls it Experientialism—any proposition for which I enjoy this ex ante, basic justification is an item of my evidence. This view has at least three features that many will find attractive: (i) Unlike views that identify evidence with a certain kind of mental state or event, Experientialism entails that evidence is propositional, and hence apt to stand in relations of entailment, probabilification, and explanation. (ii) Unlike Classical Foundationalist views that restrict my evidence to facts about my internal mental life, Experientialism has it that the content of my evidence can include ordinary matters of what’s going on around me. (iii) Unlike Externalist views that entail that what my evidence is and isn’t might be opaque to me, Experientialism vindicates the internalist intuition of evidential parity between cases of veridical perception and undetectable error. But there’s a cost: The conditions of ex ante basic justification can be met when unbeknown to me the wall is actually white but bathed in red light. In this case my evidence will include the falsehood that the wall is red.

The possibility that evidence might be false can seem worrying. I hear that there is a body of evidence supporting the conclusion that P. My next question is just how strong this evidence is, and I adjust my opinion accordingly. It doesn’t occur to me to ask whether any of this evidence is true. The possibility of false evidence seems to introduce an extra variable that I didn’t think I had to worry about. Suppose I’m told that there is overwhelming evidence that P. Indeed, the evidence entails that P. On the face of it, that clinches the matter. But if it can turn out that the evidence itself is false then all that evidential support may count for naught.

Perhaps this reaction is a little shallow and that all that is called for here are modest adjustments to how we think and talk about evidence—adjustments that are warranted by the theoretical virtues of Comesaña’s account. It is not as though insisting that evidence is factive magically guards us against the possibility of grounding our inquiries in error. What Comesaña calls false evidence the rest of us might call counterfeit evidence—non-evidence which is perhaps innocently mistaken for evidence. Either way, it is something we would like to avoid but we can’t guarantee that we will. It is worth looking at the matter more carefully by considering the roles that evidence is supposed to play. Evidence is thought to be important stuff. Hume (1999) tells us that a wise man proportions his belief to it. Why? Roughly, but plausibly, because it’s our best route to the truth. Now reasoning from false premises is a lousy way to get to the truth. Does the possibility of evidence being false diminish its value? It turns out, I think, that this kind of worry might not plague Comesaña’s position. Although embracing his view in a way that avoids the worry may give us more than we bargained for.

2 Blindspots

The first thing to note is that from the first person point of view, concerning my current evidence, the threat of false evidence doesn’t seem to arise. Like false belief generally, false evidence will be a blindspot for me (Sorensen, 1988). What I’m justified in believing I’m justified in believing to be true. Indeed, on Comesaña’s view I am justified in assigning credence 1 to those propositions that constitute my evidence. So while E’s falsehood may be compatible with it’s being part of my current evidence, I shouldn’t countenance that as a serious possibility by giving it any credence at all. This won’t just apply to evidential propositions considered individually. Let E₁, E₂,…, E_n each be evidence for me. Since I give each credence 1 I will be equally certain that every one of them is true, and that their conjunction is true. It is not that I can’t come to rationally doubt the truth of some E which is now part of my evidence. Let’s say that it is part of my evidence that the wall before me is red. I might notice what appear to be some red lights shining on the wall. Or perhaps I take a sample scrape from the wall and find it to appear white under different light. If something like this is sufficient to defeat my justification for the wall being red then I can rationally doubt that it is red or even be persuaded that it is not. But the moment that this defeat occurs the defeated proposition ceases to be a part of my evidence. So there is never a time at which E is an item of my evidence and yet I harbor some rational doubts about its truth.

The only way for me to remain open to the possibility of (now) possessing false evidence is by allowing for some uncertainty concerning what my evidence is. Perhaps E* is not a part of my evidence and I rationally give less than maximal credence to its truth. But might I give some credence to the possibility that it is a part of my evidence and hence that may I possess some false evidence? Even this is not clearly the case on Comesaña’s account. For reasons I will return to, Comesaña seems committed to the view that whenever it is not rational for me to believe a proposition I am in a position to rationally believe that it is not. Is this good news? On the one hand it may seem appropriate to treat the truth of my basic grounds for everything I believe—i.e. my current evidence—as a given. But on Comesaña’s view false evidence is not only possible, it surely exists (for other people and for me at other times). It may seem a little suspicious that I constantly take myself to be free of the menace of false evidence especially when my current evidence keeps changing.

3 Potential evidence

In inquiry I am concerned not only with my own current evidence but also with whatever else might be out there—evidence that others might possess, or evidence I might obtain with further investigation. Common sense suggests that it is generally a good idea to get more evidence when it’s available. To take the simplest case, I am offered a bet on whether P at some given odds. I can choose to accept or decline the bet in the light of my current evidence. Or I can choose to obtain some more evidence on the matter. Provided it doesn’t cost me anything and there is some chance that it will affect my choice, taking the extra evidence is a no-brainer. Famously, with a few idealizing assumptions this bit of wisdom is vindicated by a simple Bayesian model (Good, 1967). But the model crucially assumes that ‘obtaining new evidence’ involves updating one’s opinions only on what is true. While I might be sure that none of my current evidence is false perhaps this doesn’t extend to possible evidence that I don’t yet possess. And if I give some credence to the possibility of obtaining false evidence it won’t always be worth pursuing. To be clear, the worry here isn’t that Comesaña’s view might actually lead us to different practices of inquiry. While I might state things in terms of whether to “obtain new evidence”, the practical choices I’m faced with are things like whether to perform a new experiment or whether to take a look at something. Whether I take evidence to be factive or not, in my investigations I face some risk of mistaking the false for the true and I must factor this into my choices. Still, it might seem theoretically odd that free evidence might not always be worth having.¹

It turns out however that it is not clear that on Comesaña’s account we should even countenance the possibility that I will obtain false evidence upon further investigation. To begin with, the following can seem odd. I’m about to take a look at the wall next door to see what color it is. I give some credence to the possibility that I will obtain false evidence (e.g. that the wall will falsely appear red without any defeating evidence). Nevertheless I anticipate that if and when I do obtain some evidence I will rationally be sure that it turned out not to be false. It seems odd that I might achieve certainty that my senses are not deceiving me just by looking at a wall. (I won’t speak for him, but Comesaña’s later discussion of the Easy Rationality and related problems suggests that this position might not sit comfortably with him). So there is some pressure to say that I should be certain in advance that I will not obtain false evidence when I look at a wall.

Or consider the following. I’m blindfolded and sitting in front of a wall. An Oracle that I fully trust assures me that upon removing the blindfold the wall will appear red. (Perhaps for good measure she gives me a sneak preview by way of a vivid hallucination of the exact phenomenal character of the experience I will enjoy). I form some opinion about the color of the wall. The blindfold is removed and sure enough the wall appears just as I was certain it would. It is hard to believe that I should now be more confident that the wall is red. (I can’t imagine taking some bet now that I wasn’t willing to take before I saw the wall). Now according to Comesaña, I am justified in giving credence 1 to the wall’s being red when I look at it. So by the line of reasoning that I’m pursuing here, I should already give zero credence to the possibility that the wall will falsely appear red to me.² This will generalize of course. We have come to the conclusion that I should not only be sure that I have no false evidence but that I’m not going to get any either!

4 Evidence of evidence

Evidence of evidence is evidence. It has proved tricky to pin down what this means with precision but there is clearly some truth behind the slogan.³ I see you carrying an umbrella. That’s evidence for me that you have some evidence that I lack that it will rain. And that is evidence for me that it will rain even if I don’t know what your evidence is. The matter is messy though. Perhaps I’ve seen that you’ve been talking to Juan. I know but you don’t that while Juan is a genius at predicting the weather he is also compulsively deceptive and always finds a way to cherry-pick some misleading evidence and present it to others in order to lead them astray. So that you are carrying an umbrella is evidence that Juan gave you evidence making it likely that it will rain But that Juan gave you such evidence is for me—knowing what a trickster he is—evidence that it will not rain. Here is one way of precisifying the slogan that seems to get it right.

Pooling: Let E be the proposition that the result of pooling my evidence with yours results in a higher probability for P than conditional on my current evidence alone. E raises the probability of P relative to my current evidence.

By pooling I just mean that we take the conjunction of all the propositions that constitute my evidence with those that make up yours. The trouble is that on Comesaña’s account this conjunction can result in a contradiction. My evidence might include the proposition that a wall is red. Yours might include the fact that it is white but has red lights shining on it. Conjoining our evidence will result in a contradiction conditional on which no probabilities are defined.⁴ Perhaps there is some other way we can understand evidence pooling that will vindicate the principle while being consistent with Comesaña’s position. But there is work here to be done.

Here is another very restricted principle in the same ballpark that seems right.

Deference: Suppose I’m certain that you began with the same priors as me, you always conditionalize on your total evidence, and you have all of my evidence plus more. Then I should fully defer to your credences. That is, P_M(p | P_Y(p) = x) = x, where P_M and P_Y are my credence function and yours, respectively.

As stated this would seem to apply only to highly artificial cases which would never arise. (You and I will never share that much total evidence). But we can restrict it to relevant priors and evidence in realistic cases. For example, you and I have the same information about the cards in a shuffled deck. So we start with the same priors concerning say the presence of a King in the last ten cards. We’ve each seen several cards drawn from the deck. But you’ve privately selected a few more. It makes intuitive sense that conditional on your credence being, say, 0.7 that the there is a King in the last ten cards, my credence should be the same even though I don’t know what that extra evidence is that you have. That after all is what I judge my credence would be (other things being equal) if I were to be privy to the extra information you possess.

As with the case of the value of obtaining evidence, there is a simple Bayesian model with a few idealizing assumptions that vindicates this bit of common sense.⁵ But once again, the model relies on the assumption that updating on evidence involves conditionalizing on the truth. Once we allow that someone may have false evidence, the simple principle above need not hold and the matter of deference becomes more complicated.⁶

How worrying is this for Comesaña’s view? Well perhaps we should extend the reasoning above. We concluded that while it is possible for evidence to be false I should be sure that not only is none of my current evidence false, I will not obtain any false evidence if I investigate further. It would be odd to treat myself differently than others in this regard. Surely I should only give more credence to your having some false evidence than me if I have some reason to think that you are more prone to such mishaps or the like. Now if I’m entitled to be certain that I have no false evidence then I will have to extend that confidence to you also. There are limits to this of course. I can have evidence that there are red lights shining on the wall but know that you are oblivious to them. I will thereby have reason to think you have some false evidence. This will be so whenever I have information such that were you to be privy to it some of your evidence would be defeated. (In such cases, Deference will not apply as I have evidence you lack). Setting such cases aside, we seem to have reached the conclusion that I’m entitled to be confident that while false evidence is a thing, I run no risk of having my inquiries infected by it, either in my own investigations or my reliance on others.

At this point we might really start to worry that this is too good to be true. Sometimes things are not as they appear, even when we have no reason to suspect so. The scenarios that Comesaña takes to involve false evidence will not be terribly uncommon. It does not seem realistic to be so confident that they won’t make for epistemic trouble.

5 Transparency of evidence

I’ll finish by briefly looking at what got us to the prospect of false evidence in the first place. Comesaña does a nice job of bringing out why the standard rival views of evidence have implausible consequences of their own. There is a natural rival to Comesaña’s view which accepts the propositionality of evidence and that evidence may include external world matters like the color of walls but unlike Comesaña’s view entails the factivity of evidence. This is of course the view that my evidence is simply whatever I know, or what Comesaña calls Factualism (Williamson, 2000). In crude outline the key argument against Factualism goes something like this. In standard Good Case/Bad Case perceptual scenarios it is perfectly rational to act as though you are in the Good Case even if you are in the Bad Case. It is only rational to act on the assumption that P if it is rational to believe that P. But according to Factualism it is not rational to outright believe that one is in the Good Case when one is in the Bad Case. Hence Factualism is incorrect. The key idea is the familiar thought that the norms of rationality need to be action-guiding. In the Bad Case I have no way to tell that I’m not in the Good Case. So Factualism has it that I’m rationally required to refrain from believing something and acting upon it (that I’m in the Good Case) even though I can rationally have no idea that this is required of me. Comesaña suggests the following principle to capture the idea.

Access Principle: If condition C makes it the case that it is irrational for S to believe that P, then it is rational for S to believe that C obtains.

I confess I’m one of those who feels the strong pull of such internalist motivations even if I think they need to be questioned. The Access Principle commits us to a lot. Comesaña notes that the principle as stated might be too strong and that some restriction on condition C may be required. But it is not easy to see how a motivated restriction might go. The motivating idea is captured in the slogan that the norms of rationality should play a guiding role. The thought is that in order for the requirements of rationality to guide me in what to do or think I need to be able to form rational opinions on what is required of me. It is hard to see how this can admit of exceptions. Once we start saying there are some conditions on rational belief that may be beyond our ken we seem to have given up on the idea that rational requirements must (always) be guiding. What we really seem to be committed to is the following.

Let R $\phi$ =_df. It is rational for S to believe $\phi$.

$${\mathbf{R}}\neg {\mathbf{R}}\,: \, \neg {\text{R}}\phi \to {\text{R}} \neg {\text{R}}\phi$$

This is naturally paired with the positive thesis:

$${\mathbf{RR}}\,:{\text{ R}}\phi \to {\text{RR}}\phi$$

(It is hard to see why we would balk at RR once we’ve endorsed R¬R).

I will end with a puzzle that casts doubt on such principles. Comesaña plausibly insists that it is possible for a belief to be rational and yet false. Suppose I’ve come to believe on the basis of strong evidence that Carolina is an infallible Oracle. She tells me that my current epistemic predicament is such that with respect to p, what it is ex ante rational for me to believe and what is the case come apart. That is, it is rational for me to believe p if and only if p is false. (She makes it clear that that is how things are for me now, having just heard all that she has told me). On the one hand there is nothing odd about this state of affairs. It is common enough for a proposition to be true but for me to lack sufficient evidence to rationally believe it. And if, as Comesaña insists, rationality is not guarantee of truth, it is possible for me to rationally believe p even though it is false. But being told that that one of these is my current predicament puts me in a weird position. If I believe Carolina, what am I supposed to believe about p? I should believe it only if it’s rational to do so. But if I judge that it is rational to believe it I must conclude that it is false. Perhaps then it can’t be rational to believe it. But if that’s what I think then I’ll have to conclude that it is true. Weird as this may seem, it can be hard to deny that I should take Carolina’s word for it. She has made many claims of that same form (R $\phi$ $\leftrightarrow$ ¬$\phi$) with respect to me and others in the past. And in each case we have been able to eventually verify that she has been correct. So I have strong inductive evidence that she is right on this occasion. It would be strange to simply reject what she has to say this time. So let’s suppose that it is rational for me to believe her. That is,

• (1) ${\text{R}}({\text{R}}p \to \neg p)$

• (2) ${\text{R}}({\neg \text{R}}p \to p)$

Now let’s add the two Access Principles:

• (3) ${\text{R}}\phi \to {\text{RR}}\phi$

• (4) $\neg {\text{R}}\phi \to {\text{R}}\neg {\text{R}}\phi$

And just two more:

• (5) ${\text{R}} \phi \to \neg {\text{R}}\neg \phi$

This is a modest anti-permissivist principle ruling out that it can be rational to believe $\phi$ but also rational to believe ¬$\phi$. (Comesaña himself expresses sympathy for a much stronger form of Uniqueness). And finally a Closure principle:

• (6) $[{\text{R}}\phi\, \& {\text{ R}}(\phi \to \psi )] \to {\text{R}}\psi$

Throw these together and you have a contradiction.⁷ Since (3) is at least as plausible as (4), accepting (4) commits us to the position that it is not rational to accept Carolina’s testimony. I suppose that might be right but it is a surprising result.