Knowledge is only observable indirectly
Learning and knowledge cannot be observed directly. No one can look at you standing still and tell, “You know X”. For any observation of knowledge about X to be possible, you must perform X.
Unfortunately, the performance of X is impractical in most settings. We need to rely, instead, on further abstractions of someone knowing X.
Diplomas, for example, are very useful shortcuts for society. They enable us to operate at scale — imagine how less efficient the world would be if there were no quick way to check if your doctor studied medicine. Despite their vast usefulness, keep in mind that diplomas are not perfect. They are not actual skill, nor should they be the only goal in education.
On its most fundamental level, learning seems to happen via modifications of the brain circuitry — but we cannot really inspect what is going on in there. As a consequence, learning and knowledge are not directly observable.
We cannot observe the precise modification process taking place in the neuronal networks involved. Nor can we tell how the I-have-just-learned-X state of the brain looks like once it is reached.
In other words, no one today can look at you standing still and tell, “You have learned X”. There is no machine and no human being that could remotely tell any of that.
Having learned something is a latent property. We believe it must be there, but we can only be sure indirectly. One must perform the learned thing for a verification to be possible.
This need for actual performance has profound consequences in society. For instance, in many settings, it may not be practical. The performance might take too long or put people irresponsibly at risk. This is why in modern education there are a myriad of ways of testing if learning seems to have occurred.
In fact, diplomas and accreditations exist to make the task of indirect verification even easier. They attest for the world that some amount of learning could be expressed by the beholder in a sequence of mock tests at given points in time.
Despite being powerful abstractions that help our society operate at its current scale and speed, we must keep that in mind that any abstract, indirect verification cannot be perfect.