Not sure where you're going here with your primate in a cage example. There's no real reason why a primate could knowingly be in a situation where they'd like to gather data on their Xing habits over time, to estimate an upper bound for how many X's they may commit in light of a potential (and now (over)-estimated) punishment.
Let's turn this around. I like the idea of micro-transaction pay-per-read articles. I tend to read ~0-100/day. I'd like to know, roughly, what the most is I'd spend in a month *if I were paying honestly for value I received*. Note the "honestly" and "most" there.
If, in playing with these free buttons, I read an article I like, why do you seem to assume that I'd hit "0" (well you know what I mean) or "1" in that case? I THOUGHT THE ARTICLE WAS GOOD. Here's a fake $.03. Yes, I realize this didn't cost me anything, but now I have a record of one article I would have read, which (were I being honest that day) would have cost me $.03. If I didn't like the article? *Then* they get 0.
Not all information needs to meet 100% rigorous data-quality-real-world standards to be useful. With a month of this tool, I can determine whether I'd be ballparking $5 or $50 **in a world where this existed, and I was being honest with myself**. That's it. It doesn't tell *you* anything about me, and it doesn't necessarily tell the system designer anything about my "real world" payments. But it sure does tell me about me.
TL,DR: This is an easier way to estimate a ceiling amount for how much you may spend in such a system than recording your month's browsing in an excel workbook would have been.