Two undefeated chess grandmasters meet at a chess tournament. No ties are accepted: this is a mental fight where only one can win. Each believes that they will win the match, and in fact has good reasons to believe that they will win: they each have defeated all previous games. For one of them their belief is true, for the other their belief is false, yet the winner cannot be said to know. True belief (mere opinion, doxa) does not make knowledge (episteme). We want a way to sort out true from false beliefs, and not just accidentally stumble upon truths like one of the grandmasters. To cut a long story short, we need some reliable way of sorting out true from false beliefs.
(Think of Gettier counter-examples.)
Say that one of the grandmasters looks at the chessboard and has the belief that it is 9:15. As it so happens, it is not 9:15; it’s actually 9:23. The grandmaster has a false belief. Let’s change the scenario and say that the grandmaster looks at the chessboard and has the belief that it is 9:23. As it so happens, it is in fact 9:23! The grandmaster has a true belief; however, the chessboard looks the same if it is 9:15 or 9:23. The chessboard isn’t a reliable way of sorting out true from false beliefs.
Say that the grandmaster looks at his watch during the match and the face of the watch displays “9:23”. If the grandmaster’s watch is reliable, the grandmaster looked at his watch that indicated “9:23”, and the grandmaster formed the belief that it is 9:23, then the grandmaster knows that it is 9:23.
- believe p
- p is true
- if a reliable justification, p; if not-p, then not a reliable justification.
Now for the fun bit: if your justification is reliable, then you would be an idiot not to follow it. If your reliable justification says p, believe p. If your reliable justification says p, p is true. The first two conditions for knowledge are a consequence of the third. Knowledge is reduced to reliable justification.
There must be some way to distinguish between reliable justification and unreliable justification, otherwise there is no epistemic difference between looking at the watch and looking at the chessboard.
We then assume that the watch is accurate.
If the watch is accurate and the watch displays “9:23”, the grandmaster knows that it is 9:23. Of course, a justification for the justification is a justification. If justification is required to be justified then ‘the watch is accurate’ must be justified.
We then assume that the watch has been tested. If the watch has been tested and the watch is accurate and the watch displays “9:23”, the grandmaster knows that it is 9:23.
It’s clear what the problem is: each justification requires a justification. Perhaps I am being too strict. Certainty requires an infinite regress, but if we accept something that is reasonably true, we will accept a degree of uncertainty. Unfortunately, if the grandmaster is .99 secure that a working watch will display the correct time and .99 secure that he wears a working watch on his wrist, that is .99 x .99 = .9801. Each new addition to the chain will reduce this security until the point that the difference between a justified (but not certain) belief and a guess is negligible.
Assume that we have solved this problem. On to the next one.
There are many competing foundational sources: tradition, clear and distinct ideas, religious authority, the senses, etc. We must choose a source, yet this produces a dilemma: (1) If we justify this choice in sources we do not treat it as a source; we create a new justification for the choice of sources, creating a new regress. (2) If we do not offer any justification for this choice between sources then the choice is arbitrary.
Sidestep the dilemma: use a self-justifying source. Religious fundamentalists argue that the Bible is the revealed truth from God … because the Bible says it is the revealed truth from God. This is no argument, just a conclusion/premise as premise/conclusion. This third way fails and we encounter the third point of the trident of Hans Albert’s Münchhaussen trilemma.