Against “Evidence”

There is no such thing as evidence – at least not if any genuine “evidence” would have to meet both of these conditions:

  1. It is your evidence that justifies all of your beliefs or credences that are justified at all.
  2. Evidence is at least roughly similar to what is called “evidence” in everyday English.

As I shall argue, there is nothing that meets both these conditions. There is indeed something that justifies all of your beliefs or credences that are justified at all, but it is radically unlike what is called “evidence” in ordinary English. Philosophers who assume that your evidence is what justifies your beliefs are in constant peril of being led astray by their linguistic intuitions about the meaning of ‘evidence’ in everyday English.

My conclusion will be that the word ‘evidence’ should be banished from fundamental epistemology altogether.

A.  First, let me suggest an account of how the term ‘evidence’ is used in everyday English. (Incidentally, no other language known to me has any word that corresponds closely to ‘evidence’: the French word évidence means either the property of being evident or an evident truth; German also has no closely corresponding word. “Evidence” is a much less universal concept than some philosophers suppose!)

In English, we typically use the word ‘evidence’ in conversations that aim to reach an agreed answer to a question Q. (E.g., Q might be ‘Is theory T true?’ or ‘Did the defendant commit the crime?’ or the like.) We use the “evidence” as our basis for trying to answer this question Q – where every part of what we use as our basis for answering Q is something that we are (in that context) presupposing to be a fact. We accept the claim that p is part of the evidence only if it seems to make sense to presuppose p to be a fact, for the purposes of that particular conversation.

In short, I propose, in any context C, it is true in C to say that p is part of the “evidence” if and only if p is a proposition that it makes sense to presuppose in C, given the purpose of the conversation to try to reach an agreed answer to the relevant question.

According to this account, the term ‘evidence’ is acutely context-sensitive. It usually makes no sense to presuppose a proposition that is controversial among the participants in the conversation; so what counts as part of the “evidence” usually depends on who we are talking to. The proposition that the earth is over 4 billion years old may be part of the “evidence” when we are talking to our scientific colleagues, but not when we are talking to religious fundamentalists.

B. Thus, our linguistic intuitions lead us to assume that our “evidence” consists of propositions that (i) we presuppose or believe, and (ii) are uncontroversial – part of a public “common ground”.

In my view, this explains why it seems plausible to so many Bayesian epistemologists that one’s “evidence” consists of propositions in which one has the highest possible credence, and why it seems plausible to Williamson that one’s “evidence” consists of the propositions that one knows. It also explains why so much of the literature on disagreement assumes that it is common for the parties to a disagreement to have shared evidence.

But in fact, most epistemologists do not accept that the only thing that can make your beliefs rational or justified are propositions that you believe. Most epistemologists would insist that a much wider range of mental states and mental events – including sensory experiences, sensations, memories, intuitions, intrinsically conditional beliefs, desires, emotions, plans, intentions, and the like. In general, your beliefs are not just based on your other beliefs. There are many complex and diverse processes that can lead you to form, revise, and maintain our beliefs.

It seems plausible that all of these mental states, events, and processes can be part of what makes your beliefs rational or justified. But these states, events and processes are not propositions that you believe, and they are also not public or uncontroversial or part of the shared common ground. It is in fact impossible for anyone else to share these mental states or events or processes.

So the claim that your “evidence” is what justifies your beliefs is grotesquely implausible, if the term ‘evidence’ is used in anything like its ordinary sense.

For this reason, the use of the term ‘evidence’ in much epistemology is a constant source of erroneous assumptions. We would be much better off if we eliminated the term ‘evidence’ from fundamental epistemology entirely.


Comments

Against “Evidence” — 20 Comments

  1. Hi Ralph. I was wondering why you thought the German beweis didn’t work more or less like ‘evidence’ when I stumbled on these three usage examples for ‘evidence’, courtesy of Google:

    [1] “There is no evidence that corroborates in any way the allegations of BP involvement in the Scottish executive’s decision to release Megrahi on compassionate grounds in 2009, nor any suggestion that the Scottish executive decided to release Megrahi…
Jul 18, 2010 –  William Hague –  The Guardian

    [2] “What we are emphasizing to the Pakistani government is the need to follow the evidence wherever it leads,” Rice said today in London. “This is a time for complete, absolute, total transparency and cooperation and that’s what we expect.”
Nov 30, 2008 –  Condoleezza Rice –  ABC News

    [3] Mr Cameron said: “This agreement will bring significant economic benefits to both our countries. It is evidence of our new, commercial foreign policy in action.”
Jul 30, 2010 –  David Cameron –  England’s Northwest

    I take it that [2] fits your description of common usage, and [3] would be an instance of what the French might say was in évidence. But that still leaves [1], which would appear to be good evidence (!) for ‘evidence’.

  2. In the spirit of Gregory’s comment, here are three quotes taken from three different articles:

    “A Fresno County judge has dismissed murder charges against a defendant after learning that authorities destroyed key evidence in the case…Judge Edward Sarkisian Jr. dismissed two murder charges and two attempted murder charges on Tuesday after learning that ballistics evidence, including guns, diagrams based on witness accounts and shell casings, were destroyed following the acquittal of another defendant…”

    “A BMX bicycle that a 14-year-old boy was riding when he was hit by a car in Faribault and killed has been stolen from its storage location at a towing company garage, police said Wednesday.

    The theft of the damaged evidence occurred on either July 1 or 2 from Glenn’s Towing and Service of Faribault at 628 Central Av. N., police said.”

    “A two–year–old vehicular homicide case is on hold while the Colorado State Patrol tries to search for evidence that was lost.

    Mesa County prosecutors and defense attorneys told a Mesa County judge Thursday morning that troopers botched some key evidence they needed for DNA testing.

    Both attorneys say justice is at stake. They say how evidence is handled could mean the difference between a conviction or a defendant walking free.”

    Do any of these ordinary uses of ‘evidence’ suggest that evidence need to consist of propositions?

  3. Hello Ralph,

    A few thoughts in response.

    I’m pretty sure I’d cite the fact that the Earth is billions of years old in a conversation with Young Earth Creationist.

    Also, the fact that we wouldn’t cite something in conversation does not tend to show that we’d be led to thinking that it isn’t true, or that things couldn’t be that way. I almost certainly wouldn’t say to someone, “Your outfit is ugly,” but this doesn’t tempt me to think that any person I’m speaking with must not be wearing an ugly outfit!

    Anyway, I think 2 is suspect to begin with, since there are many values for ‘F’ where it’s not true that “the Fs are at least roughly what are called ‘F’ in everyday English.” Most things aren’t ever referred to in everyday English. And, as with ugly outfits, even when they are referred to, they usually aren’t referred to for what they are.

    Also, maybe I’m just missing something, but unless we make a counterintuitive (to me) assumption, I don’t think that the combination of (i) and (ii) can really explain “why it seems plausible to Williamson that one’s ‘evidence’ consists of the propositions that one knows.” Unless it’s assumed that what one knows is uncontroversial, how could that combination explain the Williamsonian intuition?

  4. Here is an alternative diagnosis and prescription. In everyday english, evidence is called to play a number of different roles. E.g. it is called to play the neutral arbiter role and the justifying role. It might be that there is not one thing that can play two roles. That’s the diagnosis and it is stolen from Kelly’s SEP entry. The prescription is not to shun the word ‘evidence’; rather, use the word, but do not have in mind both roles at the same time. Epistemologists might need to think about evidence in the sense of justifer, but philosophers of science might need to talk about evidence as neutral arbiter. So why not simply sharpen our use of the term rather than avoid it all together?

  5. Following up John’s last comment, I took Williamson to argue in PoP that (ii) in large part explains why people have been resistant to E=K. The chapter on evidence suggests that we’re driven to psychologize the notion of evidence because we feel pushed to think of evidence as a neutral arbitrator – a ‘common ground’ – between disagreeing parties.

  6. Gregory —

    I think the German word ‘Beweis’ really means ‘proof’. (Every language known to me has some word for ‘proving’ or ‘showing’ that something is the case!) If it is true to say in German, “Es gibt einen Beweis, dass p“, then there is a way of showing that p is true. The English word ‘evidence’ is obviously weaker. There can be evidence for p even if that evidence is overridden or inconclusive.

    Martin —

    I don’t believe that your examples show that the English word ‘evidence’ ever just refers to an enduring three-dimensional object, like a bicycle or a gun. We can still interpret the word as referring to something more like a fact or a state of affairs: in some cases, the relevant state of affairs might just be the existence of the gun; but it is more likely to be something like the precise condition that the bicycle is in, or the like.

    It doesn’t seem too hard to make sense of the idea of the “preservation” of such evidence-constituting states of affairs. (Presumably, what is important to their “preservation” is their continued exploitability as evidence.) If we can make sense of “preserving” an evidence-constituting state of affairs, then it shouldn’t be hard to make sense of its “destruction” either.

  7. John —

    If you cited the true age of the earth as evidence in a conversation with a Young Earth Creationist, you wouldn’t be seriously trying to reach an agreed answer to a question with him. You would just be trying to browbeat him. That might be a perfectly sensible thing for you to do, but I do think that it would make your context extremely unusual as a context for using the word ‘evidence’.

    I actually think that Williamson could accept this point. When the creationists say, “What evidence is there that the earth is billions of years old?”, what they mean may just be: “What evidence can you give us that it’s billions of years old?” Williamson could accept that your assertion “The Earth is about 4.5 billion years old” will not result in the creationists’ acquiring that proposition as part of their evidence, because it will not result in their coming to know that proposition.

    Still, you’re right, this point by itself doesn’t show that we need a radically contextualist account of ‘evidence’ of the sort that I sketched. I would have to undertake a much longer and more detailed study of the linguistic data to show that; but I’m still confident that it could be shown!

    Incidentally, the explanation that I was suggesting of why “E=K” seems plausible to Williamson is because in ordinary English, everything that we accept as part of the “evidence” is some proposition that we presuppose to be a fact. (Williamson does not seem strongly moved by the linguistic intuitions that lead us to think that the “evidence” must be shared and uncontroversial.)

  8. Aidan —

    The facts about my mental states may not be controversial, but they also aren’t part of a public, shared common ground.

    So psychologizing the evidence is in at least some tension with our linguistic intuitions about “evidence”. (In a court of law, it is the witness’s statements that are taken as evidence, not the mental states that the witness thereby expresses.)

    Anyway, if Williamson agrees with me that our linguistic intuitions about the term ‘evidence’ are liable to lead us astray, I don’t see that that’s a problem for the claims that I was making!

  9. Ralph,

    I take your point about ‘beweis’, except to suggest that perhaps it is the notion of publicly shared and readily accepted information (evidence) that is most salient in natural language usages of ‘proof’ rather than the strictly technical notion of soundness. After all, Carnap thought that inductive support could be modeled as partial proof.

    Do you accept the usage evidence I submitted as evidence that ‘evidence’ is commonly used in English in the manner you are contesting?

  10. Ralph,

    I’m inclined to agree with you that ‘evidence’ refers to states of affairs rather than objects, but if that’s correct, does it follow that evidence consists of propositions? Perhaps talk of evidence being lost, destroyed, tampered with, misplaced, etc., suggests the following picture: bits of evidence are states of affairs, but evidence can be evidence *for me* only if I have the right sort of access to it. Furthermore, in order for evidence to be evidence *for me*, I have to grasp propositions that accurately represent the evidence.

    So consider when someone says “The evidence is staring you right in the face!” There is nothing infelicitous about this use of ‘evidence,’ and the phrase suggests that evidence can be “out there” even if no one appreciates it or recognizes it. It also suggests that evidence is “worldly” in a way that propositions are not.

    With that said, I don’t want to make too much of these examples, and I am not altogether confident that they are even suggestive in these ways.

  11. Chris —

    I wasn’t trying to show that it is impossible for philosophers to make “evidence” central to their epistemology without falling under the influence of the mistaken view that it is one and the same thing that plays the “justifying role” and the “neutal arbiter role”.

    Still, I did suggest that some super-smart philosophers have been led astray in this way (e.g. Williamson, the Bayesian formal epistemologists, half the people working on the epistemology of disagreement…). If that’s right, then surely it’s safer to avoid using the term in epistemology than simply to try not to “have in mind both roles at the same time” as you recommend!

    Gregory —

    I’m sorry, I just don’t see why you think that the Foreign Secretary’s statement cited in the Guardian (18 July 2010) is in tension with my account of the normal meaning of ‘evidence’ in English. It seems a perfect illustration of my account!

    When I said that the German word ‘Beweis’ means proof, I didn’t mean a formal proof (such as a logical or mathematical proof or the like). I just meant the “larger meaning of the word proof” that Mill refers to in Chapter 1 of Utilitarianism: “Considerations may be presented capable of determining the intellect to give its assent to the doctrine.”

    Evidence, unlike proof (even in this larger sense), may not be capable of “determining the intellect” to “assent to” anything. Even if there is evidence in favour of p, it might be outweighed by the evidence in favour of p‘s negation.

  12. Martin —

    I don’t really want to distinguish between states of affairs and propositions here. (A proposition is the object of attitudes like belief and knowledge, but I don’t see why that object can’t also be a state of affairs.)

    At all events, we have to explain how it can be the case that it is part of Lois Lane’s evidence that Superman saved the passengers, but not part of her evidence that Clark Kent saved the passengers.

    If we say that Lois Lane’s evidence consists of “worldly” entities, we’ll have to say that her evidence consists of such worldly entities “under a particular mode of presentation” or the like.

  13. Ralph,

    I like the post. Having spent most of the summer working on these issues, I guess my main complaint is that you waited until August to post this.

    You say,
    “most epistemologists do not accept that the only thing that can make your beliefs rational or justified are propositions that you believe. Most epistemologists would insist that a much wider range of mental states and mental events – including sensory experiences, sensations, memories, intuitions, intrinsically conditional beliefs, desires, emotions, plans, intentions, and the like. In general, your beliefs are not just based on your other beliefs. There are many complex and diverse processes that can lead you to form, revise, and maintain our beliefs.

    It seems plausible that all of these mental states, events, and processes can be part of what makes your beliefs rational or justified. But these states, events and processes are not propositions that you believe, and they are also not public or uncontroversial or part of the shared common ground. It is in fact impossible for anyone else to share these mental states or events or processes.”

    Seems right, but you don’t discuss what seems to be the natural view for someone to take if they think that evidence consists of true propositions/facts including some worldly facts. Why can’t we just say that the mental states, events, conditions are enabling conditions without which the reasons there are won’t justify our beliefs?

    This seems to do justice to the inchoate thought that the non-factive mental stuff plays some role in the justification of belief without depriving the non-mental, worldly facts their role as justifying reasons.

  14. If the cost of keeping “evidence” in our epistemological vocabulary is to reject “sensory experiences, sensations, memories, intuitions, intrinsically conditional beliefs, desires, emotions, plans, intentions, and the like” as justifications of belief than I’m willing to accept that cost.

    One thing (and I’m not sure how important this is to your thinking) is that evidence strikes me as very social thing (which is why context is so important). My “feeling” that Bob must have been in the room is not “evidence” of his having been there (to commit the crime) even though it may strengthen my belief in his guilt. His fingerprints (a fact about which I am able to hold “uncontroversial” true beliefs) are evidence.

    Your list of non-propositional justifications seems to indicate highly subjective factors. I don’t know if that was your intention.

  15. Ralph,

    I’m very sympathetic to the idea that nothing could meet both conditions 1 and 2. The epistemic notion of evidence is, I think, a technical one designed to play a certain theoretical role – something like that set out in your condition 1.

    We could I guess jettison the term and adopt one that’s more overtly technical – but colloquial terms are often co-opted for special, technical purposes and I’m not sure that this is any more hazardous here than elsewhere.

    And I’m not sure that the two theories of evidence that you mention – the knowledge account and the Bayesian account – really do betray the influence of linguistic intuitions about ‘evidence’ in ordinary English.

    On a simple objective Bayesian picture, for instance, the set of propositions to which I assign a maximal degree of belief *will* determine what degrees of belief I should assign across the board. That is, on this kind of picture, these propositions will play something like the role enshrined by 1 – and this, I take it, is why a propoponent should regard them as constituting one’s evidence.

    My impression is that accounts of evidence are, to a large extent, motivated and criticised according to their capacity to respect something like 1.

  16. Pingback: Evidence : Mormon Metaphysics

  17. Ralph –

    I have two questions:

    A. Following Martin Smith, I’d like to know how you deal with cases of “destroying the evidence”. It seems when evidence is destroyed, the *basis* (objects or state of affairs, say) for resolving a dispute is destroyed. It seems very odd to say that a shared belief or presupposition is destroyed. (It’s easy enough to imagine cases where no one ever sees the evidence before its destruction.)

    But if the means for resolving a dispute can count as evidence, then perhaps by extension, experiences, states, intuitions, plans, etc., can count as evidence too. All of these are (or could uncontroversially be paired one-to-one with relevant) states of affairs.

    Yes, they are, perhaps, un-shared and un-share-able. But the dispute might be with oneself, where, still, some things are up for grabs and other things are not. And these experiences, states, intuitions, and so on, could function in one’s personal investigations much like samples of blood and fingerprints in a private evidence room. (And, just as easily, get misfiled, modified, or re-discovered.)

    B. Clarificatory question. Are you claiming that only philosophers use “evidence” in accordance with (1), i.e., in accordance with the notion that all beliefs are justified (only) by evidence? Or does that use simply extend our ordinary use? That is, are you accusing philosophers of mashing up their innovated use with a previously coherent pattern of ordinary-language use, or are you finding an incoherence within the ordinary use that only crops up rarely?

    In the first case, the answer seems to be to stick to the ordinary use of ‘evidence’ and force philosophers to fashion a neologism. In the second case, we have a genuine paradox on our hands.

  18. “Thus, our linguistic intuitions lead us to assume that our ‘evidence’ consists of propositions that (i) we presuppose or believe, and (ii) are uncontroversial – part of a public ‘common ground’.”

    Regarding (i) it’s not clear that ‘presupposing’ and ‘believing’ are interchangeable when it comes to evidence. You use the disjunction in (i) as a springboard to make your later claim that, “most epistemologists do not accept that the only thing that can make your beliefs rational or justified are propositions that you believe.” This is too fast for me because it’s not clear that presupposing something as fact requires believing that it is fact. What about cases where something is presupposed for the sake of argument? The actual attitude toward the proposition may be one of withholding belief (e.g., remaining agnostic for the purpose of engaging in meaningful conversation about a particular question). It’s not clear that evidence, as neutral arbiter, must consist of “propositions that you believe”.

    Regarding (ii) I want to resist your characterization of our linguistic intuitions. What matters for presupposing a fact as evidence in a conversation is not that the agents are on the same team regarding the evidence (i.e., either naturalists or creationists). What matters is whether the agents can agree that the evidence is *relevant* to the question. Your narrow reading of our linguistic intuitions only applies to contexts of discovery. That is, when you’re helping a peer connect-the-dots. This is because your use of “belief” in (i) implies that agents must share or closely approximate the same degree of belief in the evidence in order for the evidence to be common ground in a conversation. Shared evidence doesn’t mean shared (or closely similar) degrees of belief in the evidence. If degrees of belief in the evidence must be the same in order for the evidence to be admissible in conversation, then it is possible for the degree to which the evidence confirms the hypothesis to also be uncontroversial. If you both have the same confidence in the truth of the evidence, yet you vary on the degree to which the evidence supports the hypothesis, then one of you might simply lack information or not really be a peer. To play “neutral arbiter” a proposition need not be uncontroversial in that it is presupposed or believed with the same degree of confidence; it only needs to be uncontroversially relevant in order to constitute common ground. This allows degrees of belief to vary in what constitutes the facts, and it makes for lively conversation as both parties appropriate it in trying to justify a particular conclusion.

  19. Just to note a potential ambiguity in the claim that evidence is what justifies belief. On one reading, my evidence *is* my justification or reason to believe a proposition, i.e. it’s a consideration that counts in favour believing that proposition and so it is the sort of thing that can be cited in deliberation about what to believe, etc. On another reading, my evidence is what *makes it the case* that I have justification to believe a proposition, i.e. it’s a justifier.

    So, there are two distinct roles that might be associated with evidence in the general vicinity of 1. One’s theory of evidence is likely to differ depending on which role one takes to be crucial.

    I have to say, it’s not clear to me why we need the term ‘evidence’ to draw these sorts of distinctions. So, I find myself in sympathy with Ralph: what would we lose if we banished the word ‘evidence’ from epistemology except the potential for confusion with various distinct everyday uses?

  20. Let’s’ consider the clause (from [1], above): “There is no evidence that corroborates in any way the allegations of BP involvement in the Scottish executive’s decision to release Megrahi…”.

    For the sake of argument, let Q be the question whether BP was involved in winning the release of Magrahi.

    According to Ralph,

    …every part of what we use as our basis for answering Q is something that we are (in that context) presupposing to be a fact. We accept the claim that p is part of the evidence only if it seems to make sense to presuppose p to be a fact, for the purposes of that particular conversation.

    But, the denial in [1] is not claiming that there is no basis to have a conversation about whether BP intervened in the Magrahi case; rather, [1] is saying (under our supposition) that there is no fact/proposition/item of information (we can be agnostic here) which corroborates whether BP intervened, whatever context one wishes to consider, and whether or not one wishes to have a conversation about it.

    Perhaps a consequence of [1] is that there is no p which would license an agent to assert Q in most contexts, but that is another matter altogether.

Leave a Reply

Your email address will not be published. Required fields are marked *