To say that you know something is, basically, to say that you are certain of it; in effect, you are promising that what you say is true. Knowledge is important because we want, as a society, bodies of knowledge that can be relied upon. That is why, when cause for doubt is shown to us by skeptical scenarios, our natural reaction is to doubt that we did have knowledge; although of course, academics cannot conclude that they know nothing. At the other end of the scale, consider a boy sitting an exam, who is not sure of an answer but puts it down anyway, and it turns out to be correct: we say that he did know the answer. There are a range of uses of the word "know," and in between those two are all the sciences, and all their applications, and such varied uses of "know" give it a certain inconsistency. The analytic-philosophical analysis of "know" is therefore a cornucopia of papers. Continental philosophers may notice that you can only ever really know what you have yourself made up, however, because the paradigm case of knowledge is, as it has always been, that of a God: proposition P is known by subject S when S's justification for believing P guarantees that P is true. How close you have to get to that ideal, for what you believe to count as knowledge, depends upon the kind of knowledge that it is, the use that you are going to make of it, and so on. We pick up on the use of "know" as we learn English, and I for one have found that good claims to knowledge can be gambles, akin to promises, even though knowledge stands opposed to epistemic luck. More generally, we might disagree about the meaning of "know" without any of us being wrong. But one thing stands out: you either know something or else you do not. Justification, by contrast, comes in degrees; and that is so even though knowledge is basically justified true belief. That is because the justification required for knowledge is sufficient relevant justification. Note that if you think that you know something, because of some justification, but your belief turns out to be false, so that you do not know it, then as a rule your standard for sufficient relevant justification in similar cases will need to be revised.