Originally posted by Soothfast
S may be justified in believing P despite S having reasons insufficient to guarantee the truth of P. But S can't know P under those circumstances.
Are you saying, then, that (3) can be true and yet "S knows P" false? But then, if P happens to be true, it should be possible for (1)&(2)&(3)&(4) to be true while "S knows P" is false. Thus ...[text shortened]... be false.
EDIT: I'm going to go study the "Gettier counterfactuals" more closely.
Yes, S can be justified in believing P despite P being false. In face, S can be justified in believing P, P can be true, and S can still not know P (because S's justification is not connected in the right way to the truth of P; this is what Gettier showed).
No, it doesn't follow that if
all the conditions are met, that 'S knows P' could still be false. Here is a quick Gettier example:
Suppose Smith believes his friend Jones has a brown car. He has seen this brown car, been driven around in it, etc. Just yesterday, Jones was talking at length about his brown car to Smith. In short, Smith is justified in believing Jones has a brown car. Further, suppose it is actually true that Jones has a brown car. Thus, Smith has a justified, true belief that Jones has a brown car. But does Smith
know that Jones has a brown car? It seems like it. Suppose, however, that today Jones sold his brown car. Jones then took the money and bought another, completely different brown car. Smith still has a justified, true belief that Jones has a brown car. But know it seems that Smith is only
accidentally justified in his belief. The evidence Smith has isn't connected, in the right sort of way, with the facts that make it true that 'Jones has a brown car'. Smith is just sort of lucky in still having a true belief.
What condition (4) aims to do to spell out the connection that one's reasons, evidence or justification for a belief has to have to the facts that make that belief true in order for one to have knowledge. There have been a number of attempts (e.g., Armstrong's causal analysis, Nozick's truth-tracking analysis, Clark's no-false-lemmas analysis, Harman's no-essential-false-lemmas analysis, etc.
ad nauseum); some are better than others.
You know that a proposed condition (4) is bad if you can provide a counterexample showing that each of the conditions is met and yet, intuitively, S doesn't know P. But it is not inevitable that there will always be a counterexample to any proposed condition (4). This is why it doesn't simply follow that the conditions as a whole don't provide a good analysis. Work needs to be done, though, that's for sure. And then, of course, it could be the case that the whole project is doomed because conceptual analysis never works for natural language concepts, that Wittgenstein is right about the different instances of 'know' bearing family-resemblances to each other, not sharing definitions; that Quine is right about analyticity; that Kohlberg is right about knowledge being a natural kind; that Goldman is right about knowledge being the result of reliable true belief formation, etc.