Lidless:
Thus any information is incomplete and all knowledge is an estimation of reality.
I have to agree. If knowledge is defined as perfect correlation between perception and reality, one must be able to stand outside both and evaluate their correspondence. We can't stand outside either of them.
Ax:
...but at what point is the data sufficient to call the result knowledge?
Lewis Thomas has a nice quote on this: A fact is simply the point at which investigation has ceased.
To answer your question in a word: Never.
Here's something else to cook your noodle. Not only is all knowledge a probability assessment but humans do not assess probabilites correctly. We have inherent 'cognitive heuristics' that guide our assessment of probabilities and many if not most of them are fallacious. Even statisticians invariably commit these fallacies if they are not alerted beforehand to the nature of the test.
The fact that these cognitive heuristics are so widespread suggests an adaptive advantage to evaluating things improperly, but no one has yet figured out exactly what 'formula' people are using when they commit these fallacies.
For example: the fallacy of Representativeness. This causes people to give a plausible narrative a higher probability of being true than the individual events contained in the narrative. According to the laws of probability, an individual event must have a higher probility of occuring than the joint probability of that event occuring together with another. The cold war conviction that the Soviet Union would attack us has been attributed to this cognitive fallacy on the part of strategists, who routinely drew up 'scenarios' of diplomatic deterioration and assigned war a much higher likelihood than it every really had.
Take a poll among your friends and tell them you've just flipped a penny six times and it came up heads every time. Ask them how likely they think it is that the next flip will be heads as well.
I played that game with five PhD engineers and every one of them gave the wrong answer.
Jn