[Expanded definition]

Evidence /'evədəns/ noun

1. Any observable data, rational consideration, or other information that provides a reason for believing, doubting, or disbelieving a claim within a particular domain of inquiry or practice. What qualifies as evidence can vary depending on the specific methods, assumptions, and criteria of the field in question, ranging from empirical data in the natural sciences, to logical proofs in mathematics, to personal experiences or anecdotes in everyday reasoning.

2. The available body of facts, observations, or other information indicating whether a belief or proposition is true or valid. Evidence serves as a functional link between claims and reality, helping to shape beliefs and models to be more responsive to and predictive of the world.

3. (Law) Information drawn from personal testimony, a document, or a material object, used to establish facts in a legal investigation or admissible as testimony in a law court.

4. Signs or indications of something; manifestations or symptoms. In this sense, often used in the plural (e.g., evidences of disease, evidences of poverty).

5. (Phenomenology) The immediate, pre-reflective awareness of one's own conscious experiences; the self-evident nature of phenomena as given to consciousness.

verb (evidences, evidencing, evidenced)

1. To make evident; to show or demonstrate clearly.

Updated

I mean, this is obviously wrong in all sorts of ways. We can come up with a thousand contrived counter examples where we'd say E is evidence but P(H) is 1 (or 0, given your amendment), so P(H|E) doesn't change.

@Najawin Did not see the updated one, just saw the comment. Imagine I tagged the updated one. I am good at the manifold UI. I promise.

@Najawin If P(H) is already 1 or 0, I think it's pretty plausible to say that E isn't evidence.

@PlasmaBallin I'm not aware of anyone who uses the term like this, no. The obvious area would be something like logical or mathematical derivation. If I can show that A and B imply C through one argument, if I sufficiently qualify the system I'm working in as I'm stating H, P(H) is 1 moving forward. But if I'm shown a different argument that A and B imply C in that same system, surely that's still evidence, no? Just counterfactually, if I had seen that argument first, it would have been evidence, but since I saw it second it wasn't?

And this is just a really simply example. There's all sorts of weirder ones that make this harder to maintain, with logically necessary statements and Hempel's Raven-esque approaches.

@Najawin Really good point. If we now know that something is objectively true, P(H)=1, that doesnt mean that anything which supports that hypothesis is not evidence.

I think the above clause with the counter clause would bring us close to the meaning of evidence. it can be evidence if P(~H/~E) increases compared to P(~H).

@HarishGanesan I think this is still weird though when it comes to math or logic. If some argument, E, fails to establish H, well, that, in and of itself, doesn't really change my credence for H, does it?

There's larger mathematical tool kits we build up for how to approach proofs, the role of intuition in whether we think things may or may not be true based on how "mathematical objects react to our prodding" as we try to prove things, etc etc. But there's a holistic epistemological framework here that we're dealing with in order to come to our beliefs about whether unproven mathematical statements are likely to be true or false. You can't strip it down to just one proof failing.

This one proof succeeding would be evidence of H. But in and of itself its failure is not ever going to be evidence that H is false, you need much more robust machinery here.

(And this is why I think "that which justifies belief" is the best answer here. Because it defines evidence in terms of two of the atomic terms in epistemology, belief and justification, and then gets completely out of the way, not taking up any more cognitive space, letting us get onto the hard problem of dealing with what it means for beliefs to be justified. Which is what we care about in the first place, tbh. Also it's what the SEP uses. So that's just a good reason to pick it as a minimal definition.)

0 and 1 are not valid values for priors.

It could always be the case that, say, a Cartesian demon is the source of all your perceptions, or something like that. The probabilities for this kind of thing is minuscule, but not zero, so if you're assigning probability 1 to anything you're failing to account for that. That's not an issue with the concept of evidence, it's an issue with how you're using it.

There are necessary facts that you can know a priori, even a Cartesian demon can't sway you.

Eh. Not so sure. Even those still seem contingent on a choice of tools to look at the world. They are useful tools, we are in agreement that we should use them - but that doesn't make them less of tools on which these "a prior" truths are contingent.

But this just pushes the issue back one. If you think that Q isn't itself necessary a priori, but is contingent on P, then it's necessary a priori that (P->Q). So the fundamental point I'm making is still true.

If you think that Q isn't itself necessary a priori, but is contingent on P, then it's necessary a priori that (P->Q).

This argument is itself one such tool. Like I said, I agree with the tools you use, I use them myself, and they are useful; but they're not in and of themselves some inherent part of the world we observe, and our activity if trying to make sense of that world is where probabilities and evidence live.

This itself does not work. Because you can take the conjunction of all such arguments. This is necessarily then necessary a priori. There's a reason practically nobody accepts epistemic anti realism. It's obviously and trivially wrong. People really only default to it when they want to escape another conclusion. (See here either Fear of Knowledge by Boghossian, or, a bit more round about, but has a substantial section dedicated to this, and is, imo, better, The Normative Web by Cuneo.)

Sure sure

0 and 1 are still not probabilities

What I just described generates a probability of 1, because it's necessary a priori.

👍

any market about a poll can never be as good as the poll

just make the poll

@Jono3h But with a poll you cannot have pollees add options while it's ongoing (I think - or if you can, it's worse than getting all the options beforehand, because anyone that voted before an option was added didn't get a chance to vote for that option)

This should be "if the posterior probability P(H|E) is greater than the prior P(H)", otherwise it's evidence against H

@JakobBrunker Oh, when I said "evidence for", I guess I meant to say "evidence relevant to" (i.e., it was not supposed to mean "in favor of"). Maybe I will submit a corrected version that is clearer.

Bad definition. Evidence may compel a rational agent to believe one statement over another, but it only compels one to change one's course of action in the special case where the evidence favors the hypothesis that one course of action is better than others. There are lots of things I could receive evidence for that wouldn't compel me to prefer one course of action over another.

@PlasmaBallin

There are lots of things I could receive evidence for that wouldn't compel me to prefer one course of action over another.

"A difference which makes no difference is no difference at all", or so say the pragmatists like William James. But I agree "course of action" is poor phrasing of the idea, since it implies the action needs to be realized, instead of simply planned for.

Gettier condition? Never heard of it.

You can have some evidence for a proposition and nevertheless not know the proposition. Smith clearly has evidence for his belief, despite the fact that his belief does not qualify as knowledge.

Oh I know. The comment is more about finding a definition for evidence that tries to reduce it to primitive terms in epistemology.