When a measure becomes a target, it ceases to be a good measure.

The more you prepare specifically for a test, the harder it is to determine how much of your score is based on your general understanding of what is being tested, and how much is based on you exploiting the test.

But it seems humans only tend to acquire and demonstrate knowledge for some sort of self gain. These tests would then just tell you the general level of apathy of the students and do not tell you how much they know.

If I wanted to know how well the average person on the street can clean a toilet and my test involved asking a bunch of people for a demonstration without pay nor feedback of any kind I don't suspect I would find a lot of clean toilets. I would instead have learned that people don't like cleaning toilets and gained no insight into their ability to do so.

And the more you discriminate between good students who prepared the test and the bad students who didn't bother, exaggerating the contrast.

If I don't study at all ever, only then is my understanding pure and not contaminated by studying for a test?

If I dont study for the test, then you basically measure how long ago it was when I was revising the material last time (if it was last week I get much better result then if it was month and half ago) and you have even more random in terms of whether my learning coincided with I focused on what would count as "exploitation" of this test or different test. For example, was you focusing on speed during easy exercises or ability to solve hard problems last three weeks? It is pretty much random choice and preparation lets you get back to other mode relatively fast - while without it you get worst results with kids who accidentally did wrong thing and it has nothing to do with understanding of material, just a random chance.

Motivation to actually prepare yourself removes that random.

The exploitation of specific test, especially if you dont understand base material, gets you only so far compared to that.