How should we respond to the challenge from Gettier? I think there are essentially three options.
Add another condition to the three conditions in the tripartite analysis that deals with luck better.
Get rid of the justification condition and replace it with a better anti-luck condition.
Develop an understanding of justification which successfully does the necessary anti-luck work.
Reliabilists, of the two different varieties we’ll discuss, take options 2 and 3 in responding to Gettier.
Taking option 3
By taking option 3 they maintain the tripartite analysis of knowledge, but define justification as follows.
Reliabilist Justification: S is justified in believing that p if S’s belief was the product of a reliable belief-forming mechanism.
Taking option 2
Alternatively, the Reliabilist takes option two and defines knowledge (as opposed to justification) as follows.
S knows that p if:
The overarching idea behind both accounts is that lucky guesses are not knowledge because they are unreliable – by guessing you get lucky some of the time, but guesses hardly ever produce true beliefs. In order to be justified, or to know, you need to have formed your belief in a way that will make it probable that it is true.
The trouble with both kinds of Reliabilism, however, is that reliability is not the same thing as infallibility. And so it is possible, under Reliabilism, to have a belief that is the product of a reliable mechanism, but nevertheless false. As such, Gettier scenarios will always be possible in theory – situations where a belief that was the product of a reliable belief-forming mechanism would normally be false, but by chance tuned out to be true, and so, under Reliabilism, knowledge.
One may instead be inclined then to take option one above. For example, one might improve the tripartite analysis by adding the following fourth condition:
The trouble with this account is that it makes S rather infallible – S can only know that p if S could not possibly have made a mistake as regards p. According to this definition of knowledge, knowledge is much more difficult to achieve than we ordinarily think. This makes the proposal counter-intuitive and possibly risks committing us to scepticism, the view that we know nothing at all (or at least, much less than we ordinarily think). If we need to be infallible in order to know, then, given the fact that we, humans are fallible it seems like we do not know much at all!