Justified True Belief

In the early 20th century many philosophers believed that knowledge was 'justified true belief':
A person S can consider a proposition P to be knowledge if, and only if:

1. S believes that P is true
2. P is true
3. S is evidentially justified in believing that P is true

Not long after this definition was proposed Edmund Gettier poked it full of holes by providing counterexamples demonstrating conditions that satisfied the three conditions, but that could not be considered knowlede:

Smith enters a room and seems to see Jones. He immediately forms the justified belief that "Jones is in the room". In fact, Smith did not see Jones, but rather, a lifelike replica. However, as it happens, Jones is in the room, though Smith has not seen him yet.

Smith has a justified true belief that few observers would consider to be knowledge, although it does satisfy the conditions above. A number of modifications of the JTB scheme have been proposed, and yet none have really solved the problem. It remains one of the most important questions of modern philosophy.

The obvious problem with any variation on JTB is the 'T' part. Any evaluation of 'Truth' must be made from the point of few of an omniscent observer. This makes the scheme completely useless for me, as I am clearly not omniscent, and therefore I cannot evaluate the conditions. I could ask someone else who might know, but the problem remains, I have know way of knowing if they are mistaken (or willfully misleading me), or if my preception of their answer is correct.

Likewise, the 'J' part is difficult. Without the 'T', justification isn't absolute, its just 'probable'. So what we call 'knowledge' is reduced to things that I have some good reasons to believe. Clearly the reasons have to themselves have some good supporting reasons.

Circular, or more precisely, interwoven supporting reasons are the only way to avoid an infinite regression of reasons. Since such a network of supporting facts could be internally consistant and yet still be wrong, it is important to compare the system to observed reality. Those places where large areas of the system have no directly observable counterpart in reality should be questioned. Thats not to say one shouldn't make some guesses as to how things might link together, but such conjecture should be considered fiction until it can be supported via experimentation or direct observation.

In practice this is probably what most people actually do. The differences come about in what people consider valid support. For example, someone of faith might consider a particular feeling they experience when praying over a question to be valid, repeatable (more or less) empirical data that supports the idea that prayer is a valid method of gaining knowledge, and there there is an entity that delivers the answers. Someone like myself would consider that to be (exceedingly) weak evidence. It would be just as valid to conjecture that it is a method of making a random decison and feeling good about it.

The mind is, at this point in our science of the mind, too opaque to make strong statements about why we might feel one way or another when meditating/praying on a given question. This does not mean that the technique is not useful. Indeed, having a simple way to gain confidence and happiness without actually doing anything but sitting quietly for a while is a great thing.