Gettier certainly knew what he was talking about, if he was talking about logic. The question is, how far is formal logic relevant to this. IMO, formal logic contributes to the muddle and confusion that surrounds this.

Of course Gettier problems are real. The question is whether they have a solution.

If you can come up with an example in mathematics of a Gettier problem, then you will have demonstrated that his observations apply to mathematics.

I hear it's a 2.5k year old definition attributed to Socrates, no less. :chin: — Agent Smith

I'm afraid Socrates/Plato rejects the JTB definition. Check out the Theaetetus.

The target proposition is always false, one of which the believer cannot possibly be justified in believing. — creativesoul

The target proposition in the farmer example is "There is a cow in the field" and the story tells us that there is a cow in the field. How is that false? However, it is true that the farmer is not justified in believing it. But Gettier has an argument that he is justified in believing it nonetheless, so you need to show that argument is invalid. You are advocating a version of the "no False Lemmas" reply, which I agree with. I'm not clear whether you agree with my argument for that reply and it would be interesting to know whether you agree or have a different argument to refute Gettier's argument.

How to determine whether you were or were not "really" justified in believing P1 then? — Janus

There's no alternative to gathering as much information as you can and then deciding whether the failures were few enough to count as exceptions. There could not be a determinate answer to this, so the justification would be partial. So you could get it wrong and still be justified. That makes Gettier cases possible. (Actually, the doctor is almost certainly in the same situation, that the tests and evidence will only give their answer on the balance of probability.)

In this case, it's the cow that he saw that establishes his conclusion that there's a cow in the field. He is mistaken about which cow he saw, but that doesn't undermine his conclusion. — Andrew M

In the presented case, the farmer does misidentify the cow as Daisy, and so thinks that Daisy was the cause of his perception. — Andrew M

I'm sorry, I thought the causal theory of perception was about what actually caused the perception and that what the perceiver thought was the cause was not relevant. That's a very different theory.

I'm sorry, I just posted in reply to you but omitted to include a "reply to"

But let me add a comment on

In this case, it's the cow that he saw that establishes his conclusion that there's a cow in the field. He is mistaken about which cow he saw, but that doesn't undermine his conclusion. — Andrew M

Yes, I think I contradict myself in my account of this.

So you could get it wrong and still be justified. That makes Gettier cases possible. — Ludwig V

But what does “justified” mean here? If “justified” is a normative term and not just a descriptive term, then justification must refer to some cognitive processing assessment wrt some cognitive normative standard (e.g. deduction laws in case of justification based on deduction, observational/measuring standards in case of perceptual justification, communicative standards in case of third-party feedback justification etc.)

(Actually, the doctor is almost certainly in the same situation, that the tests and evidence will only give their answer on the balance of probability.) — Ludwig V

I was thinking about probability too in reference to partial justification. But then I spotted 2 issues: 1. as far as I know, Gettier’s examples do not talk about beliefs in probabilistic terms (“S believes that P” and not “S believes that probably P”) 2. Probability either is conceptualised as a scalar value to be quantified (then what is supposed to be the probability that the farmer saw a cow while watching something that looks like a cow to consider his belief that there is a cow partially justified? I don’t think anybody is computing probabilities to support justification assessments in ordinary contexts), or it simply expresses a personal degree of confidence, but the inconvenience of taking into account degree of confidence in assessing partially justified beliefs is that anybody could be claimed to be partially justified in believing literally anything to be the case (including contradictions!) on condition that she be not sure about it. That's too much of a concession to me.

That's why I think that talking about partial justification makes more sense in ordinary contexts as a way to acknowledge some limits in our cognitive competence.

In this case, it's the cow that he saw that establishes his conclusion that there's a cow in the field. He is mistaken about which cow he saw, but that doesn't undermine his conclusion. — Andrew M

One point about the analysis I was offering is that it is deductively more palatable. Whenever we've talked about Gettier on the forum, or introduction is a real sticking point, and thus existential generalization is. People accept it in math class, but they balk at someone in real life inferring something of the form A v B v C v ... from A. I went around that by treating the name as a description. Now, instead of or introduction, we have and elimination, which doesn't seem to bother anyone.

Or introduction is not a crucial element of Gettier cases; it's just the easiest way to construct them. All we need is a situation in which your reasons for believing some proposition are not the reasons it's true. Stated abstractly, I think it's obvious this happens, and that when it does we think of this exactly as being lucky, a little like this:



"It's Disembodied Reggae Space Voice, but that's just a coincidence, you didn't know that!" (If you let the clip play on, you'll also be treated to Phineas asking Baljeet to quit arguing with the soundtrack. That show ...)

Descriptivism as a theory of names is controversial, of course, but there are a couple of specifics here in its favor: first, we're not looking at reference in general, but at recognition of a particular we are familiar with (to continue RussellFest, a particular we know by acquaintance); the other point is that the part of the description we keep is the sortal.

I think it is plausible to think of recognition as inherently a descriptive enterprise, involving a list of criteria. And sortals always play a special role. That post looks a little odd:

(1) I believe that's Daisy out there.
(2) I know Daisy to be a cow. — Srap Tasmaner

It's like that because what I thought I was going to do was have the farmer form a belief regarding the particular, Daisy, non-descriptively, and then infer further beliefs from his beliefs about Daisy; doing that would jam an and in between the recognition and the other inferences, creating two new scopes and allowing us to screw around with the reasons they're true. (I had a sort of Twin-Earthy idea that Daisy might turn out to have been a schmow all along, but with all the other things the farmer knew about Daisy still true.)

But non-descriptive recognition is so implausible, and implausible in particular if you give up the sortal. I think our beliefs are in almost every case centered on sortals; Daisy, for instance, is not just a particular, she is a particular cow. Certainly for the task of recognition, the list of Daisy's properties is going to begin with cow, and then include features (the nick in her ear, her weight, etc.) that distinguish her not from the farmhouse or the farmer's wife or the Milky Way or the tractor or democracy, but from other cows.

So all of that is to bolster the sense that you do have knowledge if you infer that you've seen a cow from your belief that you've seen Daisy the cow, even if you actually saw Clarabelle the cow.

Back to Gettier. Must there always be a false lemma when your reasons for believing a proposition are not the reasons it's true? No, obviously. It was an act of the Kansas state legislature (I'm guessing) that made Topeka the capital; you believe it because you learned it in school. There is, doubtless, a causal chain between that session of the Kansas state legislature and what your teacher told you or you read in a textbook, but it's a causal chain you cannot possibly be familiar with from beginning to end.

Now suppose you're in grade school and your teacher — because he's a bit of a prankster, or because he wants to make some point about remembering, or whatever — writes a list of cities on the board and a list of states. His intention is that he'll catch out some of the students wanting to match up Wichita to Kansas, even though Wichita is not the capital, and both Wichita and Kansas should remain unmatched. But, because he's also vulnerable to accessibility bias, he actually writes Topeka on the board. When he asks the payoff question, "What about Kansas?" a bunch of kids say "Topeka!" and he responds, "Oh-ho! But Topeka" — here he looks at the board — "is right there. Shoot." He intended there to be a conflict between the list the students had memorized and the list in front of them, to see if they could be tricked into taking what's in front of them instead of relying on what they remember, but he inadvertently made the lists the same. Now he has no experiment, because some of the students may have done exactly what he hoped, chosen Topeka remembering only that it's in Kansas. That's necessary but not sufficient for being the capital of Kansas, so it's not wrong, but it's still a mistake. But because of his mistake, it's a mistake that's undetectable. Of course, little kids tend to be pretty candid, so if he just asks, "How many of you remembered that Topeka is the capital of Kansas?" and "How many of you chose it because it's only city in Kansas on the board?" he'll probably get some hands up for each, and some embarrassed giggles.

And there's the other part of the Gettier case. Our epistemic agent always has explicit knowledge of what reasons they're relying on, what they're inferring from. To defeat no-false-lemmas, we have to construct a case in which those reasons are true and do provide strong enough support for the conclusion, but we can't do that counterfactually — that is, with reasons that the conclusion might have been true but isn't — so what we need are independent reasons, as "I learned it in school" is, epistemically if not causally, independent of "The Kansas state legislature said so."

My classroom wasn't intended to be a Gettier case, only a neighbor that illustrates the issues. But it's close, because it has elements of getting the right answer for the wrong reasons, only it adds a twist that the wrong reasons are coincidentally the right reasons. That's a funny thing, because it's almost as if "Wichita" is misspelled "Topeka" on the board, but in the teacher's mind is still the word "Wichita". So there's a false lemma here, but pushed back from the kids to the teacher. Never even realized on the board, but it's there in the teacher's beliefs. It's similar to my thing about Russell's clock, having the worker set the clock correctly from a watch that only happened to be right. Pushing the false lemma out over the agent's epistemic horizon leaves us in an uncertain position I think: depending on how the details are presented, the agent might strike us genuinely lucky to acquire knowledge, or too lucky for his belief to count as knowledge. It's like the conflicting intuitions among philosophers about the fake barn cases. What's interesting here is that most of the kids can probably report whether they had knowledge because they know whether they remembered, and it's the remembering that would be factive, but even some of them might not be sure they would have stuck with Topeka had Wichita been written on the board. And some might not know whether they would have remembered without being prompted. "I know it when I see it" is a real thing.

I'm going to take a break, but I really think we should be able to construct a clear case, roughly along the lines above, of Gettier case without a false lemma.

The target proposition in the farmer example is "There is a cow in the field" and the story tells us that there is a cow in the field. How is that false? However, it is true that the farmer is not justified in believing it... — Ludwig V

Yes. My mistake there. I was irritated at the time by another posters' hubris, very tired, and was not thinking clearly. The target propositions are true, not false. The beliefs are all false, not true. The propositions are not equivalent to the beliefs. The basic point I'm making is that S's belief is not being properly taken into account by any Gettier case, and that's the fatal flaw of them all, despite the fact that there are remarkable differences between Gettier's paper and the cottage industry that followed.

...But Gettier has an argument that he is justified in believing it nonetheless, so you need to show that argument is invalid... — Ludwig V

That's not the only way to show how Gettier is mistaken. Gettier's logic is impeccable. However, an argument can be both impeccable and false. In Gettier's paper, the fatal flaw is treating Smith's beliefs as though they are equivalent to the naked propositions he discusses. They are not. I can and have shown how that is the case.

...You are advocating a version of the "no False Lemmas" reply, which I agree with. I'm not clear whether you agree with my argument for that reply and it would be interesting to know whether you agree or have a different argument to refute Gettier's argument. — Ludwig V

I'm not even sure that I understand the NFL objection. If my answer to Gettier cases counts as a version of the NFL, then it is by pure coincidence. I'll try to adequately summarize the individual cases in this post, because the crucial parts of my view have been sporadically littered throughout my replies here in a rather disparate looking fashion. Taking the cases one at a time should clear up any misunderstandings...



In Case I, Smith is justified in believing that he will get the job and he knows that he has ten coins in his pocket. Gettier uses this justified belief and the rules of entailment for Smith to go from "I will get the job and have ten coins in my pocket" to "the man with ten coins in his pocket will get the job". Unbeknownst to Smith, another man also has ten coins in his pocket. That other man got the job. Smith did not. So... the claim is that if Smith was justified in believing P, and P entails Q, then Smith is justified in deducing Q from P and thus justified in believing Q. Q turned out to be true when treated as a naked proposition. Q is not a naked proposition. Q is Smith's belief. The difference between Q as a naked proposition and Q as Smith's belief is paramount. When Smith believed "the man with ten coins in his pocket will get the job" he was thinking about himself and no one else! The fatal flaw of the case is Gettier's failing to keep in mind Smith's belief. Smith was not justified in believing anyone other than himself would get the job. Smith was not justified in believing anyone other than himself had ten coins in their pocket. Someone else had ten coins in their pocket and someone else got the job, contrary to Smith's belief. Smith's belief turned out to be false despite the fact that "the man with ten coins in his pocket will get the job" turned out to be true when treated as a naked proposition. The truth conditions of Smith's belief do not match the truth conditions of the naked proposition. Thus, to treat Smith's belief as a naked proposition is to engage in an accounting malpractice of Smith's belief.


In Case II, Smith is justified in believing Jones owns a Ford. Gettier uses that and the rules of disjunction for Smith to go from Smith's belief that Jones owns a Ford to belief that either Jones owns a Ford or Brown is in Barcelona. Turns out that Jones does not own a Ford and Brown is in Barcelona, so again - like the first case - the disjunction is true when treated as a naked proposition/disjunction. The fatal flaw in that case is equal to the first case in that Gettier is misattributing belief to Smith by treating Smith's belief as a naked proposition when it is not. Smith believed that the disjunction was true because Jones owned a Ford. Gettier did not render Smith's belief that way. The disjunction was not true because Jones owned a Ford. Rather, it was true because Brown was in Barcelona. So, Smith's belief was false. Thus, putting Smith's belief in terms of P or Q is treating Smith's belief as a naked proposition. The truth conditions of the naked proposition are not equivalent to the truth conditions of Smith's belief. Hence, to treat Smith's belief as though it is a naked proposition is to engage in an accounting malpractice of Smith's belief.


The cottage industry repeats the accounting malpractice, but not quite in the same way as Gettier. Those cases do not follow the S knows that P formulation that Gettier addresses in his paper. Gettier gets Smith's belief right to begin with, but then conflates naked propositions/disjunctions with Smith's belief. The cottage cases begin by not getting S's belief right to start with. So, the critique/refutation for them is slightly different than the critique of Gettier's two cases. Gettier and the cottage industry all get S's belief wrong, they just go about it in different ways.


Belief that a piece of cloth is a cow does not entail "a cow is in the field". Belief that a barn facade is a barn does not entail "a barn is in the field". Belief that Clarabelle is Daisy does not entail "Daisy is in the field". Belief that a broken clock is working does not entail "it is two o'clock". Etc.

Correctly stating S's belief in the beginning marks the end of the cottage cases.

Gettier's two cases are both justified false belief. All of the cottage industry cases present true statements that do not follow from S's belief. No Gettier case offers an accurate account of S's belief.

That's why I think that talking about partial justification makes more sense in ordinary contexts as a way to acknowledge some limits in our cognitive competence. — neomac

I see your point. But there's an issue about how far philosophy needs to cater for ordinary use of words. For example, I have no doubt that someone who says "I knew that horse would win the race. A tipster told me so." did not know. That person is (misusing "know" to express subjective certainty and so undermining the distinction between knowledge and belief. But someone who claims to know on weak evidence.... that's a different issue. I don't think it is possible to develop a clear criterion.

I don't think we can escape the problem because knowledge based on statistics is everywhere in our lives.

I do think it is appropriate for philosophy to have criteria somewhat stricter than ordinary language but insisting that all justification is conclusive would result in two senses of "justification" and hence two senses of "know".

One point about the analysis I was offering is that it is deductively more palatable. Whenever we've talked about Gettier on the forum, or introduction is a real sticking point, and thus existential generalization is. — Srap Tasmaner

I don't teach logic and never have, but I do know that I had trouble accepting existential generalization, but I encountered it in the deduction from the particular to "some", and finally accepted it because "Daisy is in the field" clearly implies "There is one cow in the field" and I had to accept that "some" could include "one" at least for the purposes of logic; that then validated "There are some cows in the field". (Strictly, of course the fact that "cows" is in the plural excludes "one", so it was a stretch.) Anyway, once one has that, it is relatively easy to argue that "some" implies a disjunctive list of them.

I still feel uncomfortable with it, because there is something odd about saying that there is a cow in the field when you know perfectly well it is Daisy. One might do it if it would be a bit awkward to admit exactly what I'm going to the chemist for. But under normal circumstances, I think it is just weird. But logic doesn't take account of that.

I wouldn't discourage you from trying to construct a Gettier case that doesn't involve any false lemmas. The only one I've seen (and I've forgotten where I saw it, sadly) was clearly not a Gettier case. The problem for me is that one can construct an endless array of possibilities and lose the plot in the resulting complexity. Perhaps that's just me. I wish you luck.

all justification is conclusive would result in two senses of "justification" — Ludwig V

I don't understand what you are saying here. What are the two senses? By "conclusive" are you referring for example to sound deductions as opposed to valid deductions?
I’m still wondering what “partial justification” means. How can probability make the justification partial?

Let’s distinguish two intellectual tasks: the first one is to assess whether JTB is an acceptable definition for the notion of “knowledge”. I think that deductive reasoning offers a study case to clarify the alternatives wrt the notion of “justification”: if “justification” amounts to “sound deduction” then knowledge=JTB is still plausible (this view is in line with the NFL assumption). If “justification” amounts to “valid deduction” then knowledge=JTB is not plausible (this view is not in line with the NFL assumption). — neomac

Most knowledge claims it seems, apart from purely logical or mathematical results, are based on observation and inductive reasoning, so I am not sure where you see deduction fitting in the picture. I was trying to use exhaustive investigation as a more solid criterion to establish justification. So in the cloth/ sheep example,

I suggested that one would only be justified in believing that one had seen a sheep rather than a cloth if one got close enough to be absolutely sure ( leaving aside absurd scenarios like ' a man disguised so convincingly as sheep that it would be impossible to tell the difference' or 'brain in a vat" or 'I'm actually not awake, but dreaming' and so on.

Of course, this bracketing of radical skepticism shows that the notion of justification cannot be definitively and absolutely pinned down; there is always going to be acceptance and putting aside of some possibility of doubt and adoption of some arbitrary standard of what should be thought to constitute evidence and hence justification for empirical claiims.

I would be quite happy to dispense with all talk of knowledge whatsoever and speak instead of more or less justified belief. That would defuse all the absurd angst and wrangling over Gettier cases.

There's no alternative to gathering as much information as you can and then deciding whether the failures were few enough to count as exceptions. There could not be a determinate answer to this, so the justification would be partial. So you could get it wrong and still be justified. That makes Gettier cases possible. (Actually, the doctor is almost certainly in the same situation, that the tests and evidence will only give their answer on the balance of probability.) — Ludwig V

Yes, that's what I've been saying. I have no problem with degrees of justification, but since the conceptual distinction between knowledge and belief is very clear, and the degree of justification, except in cases where close enough observation and exhaustive enough investigation to eliminate the possibility of error is possible, is not clear, I think it would be better to speak of more or less justified belief, rather than knowledge, except in those cases which are clearly involving sufficiently close observation and exhaustive investigation..

Most knowledge claims it seems, apart from purely logical or mathematical results, are based on observation and inductive reasoning, so I am not sure where you see deduction fitting in the picture. — Janus

Logical and mathematical knowledge are based on deductions (see theoremes). And one could question that inductive reasoning based on observation can be called knowledge at all (Hume deemed inductive inference are unjustified). In any case Gettier’s examples do not seem to relate to deductions nor induction. They concern particular perceptual beliefs.
But I was suggesting to consider “deduction” as a study case for better clarifying the notion of “justification” because if “justified” is a normative term (as I understand it) and not just a descriptive term, then justification must refer to some information processing based on some cognitive standard (e.g. deduction rules in case of justification based on deduction, observational/measuring standards in case of perceptual justification, communicative standards in case of third-party feedback justification etc.). And if deduction is a form of justification, then we can easily see how our acceptance of knowledge=JTB or its rejection can be rendered in terms of valid/sound deductions. In other cases of knowledge, it’s less clear, how to distinguish valid from sound information processing.

I suggested that one would only be justified in believing that one had seen a sheep rather than a cloth if one got close enough to be absolutely sure — Janus

adoption of some arbitrary standard of what should be thought to constitute evidence and hence justification for empirical claims. — Janus

Here you are confirming that justification is a normative concept not a descriptive one, since we should use some standard to assess justification, then you are suggesting what observational protocols could be provided in order to ensure justification for perceptual beliefs (e.g. one has to go close enough to be justified). I would add that perception is not the only way we form beliefs, but there is also deduction and third-party feedback. So what is left to clarify is if there is a way to distinguish valid/sound justification in case of perceptual beliefs and socially transmitted beliefs as much as we distinguish valid/sound deduction.

In any case Gettier’s examples do not seem to relate to deductions nor induction. They concern particular perceptual beliefs. — neomac

If I see a cloth and I think it is a cow, is that not based on induction? I've seen cows before and that looks like a cow so I conclude that it is a cow.

And if deduction is a form of justification, then we can easily see how our acceptance of knowledge=JTB or its rejection can be rendered in terms of valid/sound deductions. In other cases of knowledge, it’s less clear, how to distinguish valid from sound information processing. — neomac

Right, I understand valid and sound to be two quite different criteria applied to different aspects of deductive reasoning. If my premises are sound (which means true) then my conclusion will be true provided my reasoning is valid.

It's easy enough to tell, if my reasoning is valid, not always so easy to tell is my premises are sound.

I agree that justification is a normative concept, and can be descriptive only with the context of the norms (if there be such) which are used to establish its provenance.

They concern particular perceptual beliefs. — neomac

This presupposes that there is more than one kind of belief. I find that quite germane to the topic, given the B aspect.

Regarding cases of mistaken identity...

Gazing upon a field, seeing a piece of cloth, and believing it to be a cow does not entail "there is a cow in the field". It is also not the same belief. The farmer first believed that that particular piece of cloth was a cow.<----That's the beginning of an accurate analysis of this farmer's belief. That belief grounds this farmer's subsequent thought. It also marks the end of our analysis. Belief that there is a cow in the field does not follow from belief that a piece of cloth is a cow.

No more wondering whether or not that farmer's belief is justified.

Specific instances of the Gettier problem can be generated using a template (vide infra)

Constructing Gettier Problems.

The main idea behind Gettier's examples is that the justification for the belief is flawed or incorrect, but the belief turns out to be true by sheer luck. Linda Zagzebski shows that any analysis of knowledge in terms of true belief and some other element of justification that is independent from truth, will be liable to Gettier cases. She offers a formula for generating Gettier cases:

(1) start with a case of justified false belief;

(2) amend the example, making the element of justification strong enough for knowledge, but the belief false by sheer chance;

(3) amend the example again, adding another element of chance such that the belief is true, but which leaves the element of justification unchanged;

This will generate an example of a belief that is sufficiently justified (on some analysis of knowledge) to be knowledge, which is true, and which is intuitively not an example of knowledge. In other words, Gettier cases can be generated for any analysis of knowledge that involves a justification criterion and a truth criterion, which are highly correlated but have some degree of independence. — Wikipedia

Can anyone follow the above instructions and construct an original Gettier problem, anyone?

If I see a cloth and I think it is a cow, is that not based on induction? I've seen cows before and that looks like a cow so I conclude that it is a cow. — Janus

Then there seems to be a terminological issue here. I intend inductive reasoning as a form of reasoning where from a set of particular propositions we conclude some general proposition: e.g. From crow one is black, crow two is black, crow three is black, ..., crow n is black, I conclude that all crows (in a set of crows larger than n) are black. As you see we are talking about propositions and we are moving from particular to general.
In Gettier's examples: 1. We do not talk about particular propositions as the basis for some other belief, but of perceptual evidence as the basis for perceptual beliefs 2. The perceptual belief is particular not a generalization.
So to me the information processing that goes from perceptual evidence to a particular perceptual belief is not inductive reasoning.

This presupposes that there is more than one kind of belief. — creativesoul

The expression "perceptual belief", as I use it, it's simply pointing to the genesis of that belief. If a belief is processed out of perceptual evidences, it's perceptual, if it's processed out of other propositions through reasoning it can be deductive or inductive belief, if it's processed out of a communicative channel it's a transmitted belief, etc.
This is at least part of my background assumptions while thinking about justification.

I'm not sure what the point of constructing another Gettier example would be. There are, I understand, 100 or more examples to be found in the literature already.

In any case, the template does not reflect their history (unwritten, so far as I know). You can see the changes (mostly unexplained, which is annoying) develop and evading one problem after another. For example, Gettier's own examples posit a conscious and deliberate process of inference, but later examples posit a perceptual basis, to avoid objections to that. Hence our discussion about a piece of cloth that's not a cow. But other examples posit a dog disguised as a sheep or a robot dog to get round objections - a different kind of mistake. Russell's clock and Havit's Ford try to get round mistaken perception altogether. The most recent example that I have seen posits a perfectly standard case of knowledge, which is subjected to a barrage of disinformation; "everybody" believes the disinformation, but our S misses the barrage and so only "knows" by luck. For my money, this isn't a Gettier case at all.

Long story short, after 60 years of trying no convincing Getter example has been produced. Experience suggests that they can't be. I can't help feeling that this is suggestive. Perhaps I'm not very good philosopher, but I'm inclined to predict that people are moving on.

I used to think like you, but then I looked it up in a dictionary which clearly suggested two senses of justification.

Statistics - If something is 95% likely to happen, most people would consider themselves justified in predicting that it will happen, and most people will agree.

Or consider this. The standard format for establishing who committed a crime is means, motive, opportunity. Suppose I establish means and motive beyond doubt and establish that there is no evidence against opportunity. Not quite conclusive, but enough to justify belief - or so many people would say.

↪neomac
I used to think like you, but then I looked it up in a dictionary which clearly suggested two senses of justification. — Ludwig V

Yet you claimed: "But there's an issue about how far philosophy needs to cater for ordinary use of words".

If something is 95% likely to happen, most people would consider themselves justified in predicting that it will happen, and most people will agree. — Ludwig V

So are you suggesting that the farmer calculated that the likelihood of that cow-looking thingy on the field was 95% and therefore he was partially justified in believing that there was a cow on the field? I think that’s a bit of a stretch. I could find plausible that the farmer claimed “I’m 95% sure that’s a cow” but not “There is 95% chance that’s a cow”. In other words “95%” is more likely and hyperbolically a degree of confidence not a computation of probability in the case of the farmer.

The standard format for establishing who committed a crime is means, motive, opportunity. Suppose I establish means and motive beyond doubt and establish that there is no evidence against opportunity. Not quite conclusive, but enough to justify belief - or so many people would say. — Ludwig V

That’s a good example. Would you claim that the judges know that the crime was committed based on that partial justification? Or, else, would you claim that the judges know that the crime was probably committed based on that full justification?
It seems to me that once you introduce probabilistic beliefs there is no need to talk about partial justification, the justification can still be full and conclusive (unless for you "conclusive" = "non-probabilistic" while for me "conclusive" = "sound"), the point is that premises and conclusions are probabilistic.
One way to verify this is again through deduction:
P1: if X had means and motive to commit crime Y, then it’s highly probable X committed crime Y
P2: X had means and motive to commit crime Y
C: it’s highly probable X committed crime Y
This deduction expresses knowledge not if it’s valid but if it’s sound. Contrast that to the case where the situation was exactly the same, except for the fact that the judge reasons like this:
P1: if by flipping a coin I get heads, then it’s highly probable X committed crime Y
P2: by flipping a coin I get heads
C: it’s highly probable X committed crime Y
The deduction could be valid but certainly not sound. Therefore it wouldn’t express knowledge.

My impression is that the reason why one could consider "partial" a probabilistic deduction wrt a non-probabilistic deduction doesn't depend on the lack of soundness of the deduction but on the cognitive limits that the probabilistic reasoning shows wrt a non-probabilistic reasoning in a certain domain. However since both reasoning can be sound, they both can express full justification and full knowledge.

It's true that further investigations can involve mistakes. But if we are being strict about what we will accept as believable then we should investigate as far as, and in every more thorough way imaginable, and only commit to believing when all those possibilities are exhausted. It's also true that even then we can be mistaken, but at least our beliefs would then be properly justified.

We can always resort to entertaining something for pragmatic reasons without committing to belief if we realize that our investigations have not been or cannot be, for practical reasons, adequate. So, for example, I see something moving which I think is a sheep, but there is a boundary fence that prevents me from getting close enough to definitely confirm it. — Janus

:up:

As Feynman once noted, “The first principle is that you must not fool yourself — and you are the easiest person to fool.”

Gettier cases rely on the various circumstances not being known to anyone, and so will in real life always exist unknown to anyone. As soon as they become known, they can be resolved, so I can’t see that they can be very important. Some people worry about this, but that's only because they can imagine something that's not known to anyone. That's not real life. — Ludwig V

I think that's right. The main point, as I see it, is to understand what it is that we do and why.

I'm going to take a break, but I really think we should be able to construct a clear case, roughly along the lines above, of Gettier case without a false lemma. — Srap Tasmaner

I'm not yet convinced! If the children understood that the teacher could include non-capital cities in his experiment, then I don't think the children were justified in believing that Topeka is the capital of Kansas (assuming they didn't already know it).

Whereas if the children believed (wrongly) that the teacher could only include capital cities, then that would be a false lemma. Even if, as it turned out, only capital cities were included.

It's like playing a game of chess, and finding out later that certain house rules were in effect that you didn't know about and were never utilized in the game. You "deduced" that you were in checkmate, and you were even according to the house rules, but there were more potential moves available than you thought.

There certainly seems to be a problem about the farmer’s belief that a piece of cloth is a cow. You seem to be assuming that in reporting the farmer’s beliefs, you need to use words that he would have done, and he certainly wouldn’t have said that a piece of cloth was in the field. The tricky bit is that that is exactly how we would formulate his belief and we can’t say that there’s a cow in the field (unless we are referring to the cow that is in the field, which would be very misleading.)

First off, “see” is a factive verb. In reporting what people see, we need to report what they actually see, not what they think they see. When we report what they think they see, we have to make it clear, so we need to report, not that he sees a cow in the field, but that he believes he sees a cow in the field, or that he sees what he believes is a cow in the field.

We focus too much, in these discussions, on what people say in reporting their own beliefs. But that is only one way that people show what they believe. Their beliefs also show in what they do and in other things that they say. That’s how we know that he believes that a piece of cloth is a cow. But I would use that way of putting it only to other people, not to the farmer himself. Curiously, if I was telling the farmer about his mistake, I would say “you know that cow in the field? Well actually it’s a piece of cloth.” Or “I’m afraid that cow in the field is actually a piece of cloth”. — Ludwig V

My apologies for not recognizing what all you've said here. It deserves better attention than I gave it earlier. Gestalt was in control, I suppose. I have no idea how I missed this. :smile: I wondered why you had not addressed my last reply to you, but after rereading through our exchanges, now I think I know exactly why. You had addressed my concerns(at least regarding the cottage industry cases) on a basic level here, and I somehow missed that completely, and instead summarized the basic points I've made without ever actually giving due attention to the ones you made here. Again, my apologies.

Regarding the above quote...


I completely agree that we need to report what people see. We need to report what they say. We also need to report what people believe, especially in the odd cases where they do not know what they see and/or believe about what they see. This is one such case. We seem to agree that the farmer believes that that particular piece of cloth is a cow. Where we seem to disagree is what we ought say in our report about what the farmer believes at the time. You seem to be agreeing with conventional belief attribution practices when you suggest that our report of the farmer's belief ought be what the farmer would likely say himself at that time in particular.

The farmer would not say that he believes a piece of cloth is a cow while looking at a piece of cloth that he believes is a cow. The farmer does not know that he is mistaken about what he's looking at. He does not know that he believes a piece of cloth is a cow. So, he certainly would not say that he believes a piece of cloth is a cow. That is a belief that is impossible to knowingly hold.

So, if we do as you seem to suggest here, which is in line with conventional accounting/belief attribution practices, we would not be reporting what the farmer believes. To quite the contrary, we would be going with what the farmer says at the time.

Everything can be identified under many descriptions. We use the one that is most appropriate for the context, including the method of identification that works for our audience. When we come to reporting the belief (and knowledge) of other people, we do not stick to the reference that they are using or would use; we use the reference that works for the audience we are reporting to. After all, the point is to enable our audience to understand.

It is complicated, so I hope this is reasonably clear.

Actually it's not. You've answered in what seems to be a very non-committal manner, as if straddling both side of a fence. This could be cleared up, however. I've a question for you...

Ought we report what the farmer believes(that a piece of cloth is a cow), or what the farmer would likely say at that particular time(that he believes a cow is in the field)?

A follow-up...

If we are going to go with what the farmer would say, upon what grounds are we claiming that the best time to do that(to go with what the farmer says) is when the farmer is wrong about their own belief, rather than when they become aware that they had once believed that a piece of cloth was a cow(rather than go with what the farmer would say when they're right about what they saw and what they believed about what they saw)?

The expression "perceptual belief", as I use it, it's simply pointing to the genesis of that belief. If a belief is processed out of perceptual evidences, it's perceptual, if it's processed out of other propositions through reasoning it can be deductive or inductive belief, if it's processed out of a communicative channel it's a transmitted belief, etc.
This is at least part of my background assumptions while thinking about justification. — neomac

I see.

Curious how you would answer the questions I posed to Ludwig.

I have the impression that you forgot who I am. I reviewed at length your approach a while ago and I also addressed the kind of questions you are asking now. We practically disagreed on everything.

If we are going to go with what the farmer would say, upon what grounds are we claiming that the best time to do that(to go with what the farmer says) is when the farmer is wrong about their own belief, rather than when they become aware that they had once believed that a piece of cloth was a cow(rather than go with what the farmer would say when they're right about what they saw and what they believed about what they saw)? — creativesoul

Briefly, when belief-attribution is incoherently based on knowledge-attribution, the other inconvenient is that one should update belief attribution every time there is knowledge update. The other inconvenient is that if X and Y disagree on what constitute knowledge but they agreed on what S believes according to the current belief-attribution method, yet they would report S's beliefs differently according to your belief-attribution method, multiplying the beliefs S has and probably failing to even understand each other.

I remember. That's why I asked.

I cannot make head or tails out of that answer. except that it seemed to be some sort of critique of my approach. Strange answers to very straightforward questions.

That's where you balked last time too.

Here's how I see it...

Simply put:Our disagreements boil down to the differences between our notions of belief.

I was hopeful that there was a bridge when you mentioned "perceptual beliefs", but that notion turned out to be rather empty it seems. All belief is existentially dependent upon physiological sensory perception(biological machinery), including those that are arrived at in the 'other' ways you mentioned. Thus, I found that rather unhelpful for adding any clarity.

However, that aspect, I think you called it "processing" or something similar, very well could be great material to build a bridge of mutual understanding.

That's why I asked that... first.

Strange answers to very straightforward questions. — creativesoul

Here is a more straightforward answer: we all learnt to report S’belief at t1 based on what S says at t1. That’s the practice. Now you claim that we should revise this practice because it doesn’t make sense to you for whatever reason and therefore we should attribute S’belief at t1 what S or we know at t2 about S's belief at t1. In conclusion, your belief attribution method is based on knowledge attribution. I find incoherent this conclusion because knowledge presupposes belief, so the workflow must logically start with determining belief first, and then knowledge.

I was hopeful that there was a bridge when you mentioned "perceptual beliefs", but that notion turned out to be rather empty it seems. All belief is existentially dependent upon physiological sensory perception(biological machinery), including those that are arrived at in the 'other' ways you mentioned. Thus, I found that rather unhelpful for adding any clarity. — creativesoul

It depends on what you want to clarify. I wanted to clarify the notion of justification. So to me the notion of justification applies differently based on the genesis of a given belief.

Here is a more straightforward answer: we all learnt to report S’belief at t1 based on what S says at t1. That’s the practice. — neomac

Is that what counts as a valid reply/answer these days? That may count as an answer to some people, but others can plainly see that it does not answer the questions that it should.

With regard to your question...

Indeed, it is standard practice to report S's belief at time t1 based upon what S says at time t1. That is precisely the problem in certain cases like this particular farmer story. I've shown how that practice has been found wanting, lacking, and begging for truth about the farmer's belief at time t1.

Upon what ground do you accept the farmer's self-report at time t1, when he was wrong about what he saw and believed about that, and reject his report at time t2, when he is correct about what he saw and believed at time t1?

At time t2, would you argue with the farmer about what he believed at time t1, based upon standard accounting/belief attribution practices, in the same manner you've argued against me here?

Is that what counts as a valid reply/answer these days? That may count as an answer to some people, but others can plainly see that it does not answer the questions that it should. — creativesoul

Let's try something else then: explain to me upon what grounds are we claiming that the best time to proceed our driving is exactly when green light occurs, rather than at whatever other time we feel like driving on.

I've shown how that practice has been found wanting, lacking, and begging for truth about the farmer's belief at time t1. — creativesoul

Maybe you tried. To me without success.

Upon what ground do you accept the farmer's self-report at time t1, when he was wrong about what he saw and believed about that, and reject his report at time t2, when he is correct about what he saw and believed at time t1? — creativesoul

It depends on how you construct your thought experiment: if you surreptitiously project onto your fictional character your belief-attribution method (as you did with Jack) then I would make the same objections. If he's committed to an absurd belief-attribution method that he applies to others, that method doesn't become more plausible just because he readily applies it to himself. At best, that can show that he honestly believes in its effectiveness.
If you do not project onto your fictional character your belief-attribution method , then de-re belief attribution can be successfully worked out, if there are enough contextual assumptions shared by interlocutors, but only as a tolerable derogation to the standard method of belief-attribution, not as its replacement! Indeed also those contextual assumptions are based on the standard method of belief-attribution about other interlocutors' beliefs [1]!

[1] Example: If A tells B that C believes that piece of cloth is a cow, A is (reliably?) assuming that B believes that there there is a piece of cloth and not a cow.

I think I agree with something like @neomac’s intuitions here.
Doesn’t it make sense for the farmer’s self report at t1 to be about t1 and for his self report about t2 to be about t2?
Other mental states are like that e.g. at t1 he intended to go to the shops but at t2 he no longer intended to go to the shops. (Not that at t2 he never intended to go to the shops).

his contention looks different from the one you are attributing to him.

His claims concern how we (including S himself) ascribe beliefs to some S at time t1.
Consider this case:

• At time t1 S would say "that is a cow"
• At time t2 S realizes that he was wrong at t1, and will correctly say "that is a piece of cloth"

In this case, Creativesoul would claim that we (S included) must now revise S's belief attribution at t1, and instead of saying "At t1 S believed that is cow" we must say "At t1 S believed that piece of cloth was a cow". In other words, "At t1 S believed that was cow" is a wrong belief-attribution report, while "At t1 S believed that piece of cloth was a cow" is the only correct belief-attribution report.