If I apply your logic to that, then we never know any quantity of any type of cat in any question, because there are more off-screen.
But when you are asked a question about something that is not displayed in front of you, then you do not presume that what you see is the entirety of evidence, because you are being asked about something for which what you are looking at provides no evidence. Therefore, "all cats," rather than "these cats."
"I don't see any of the blue cats, so I don't know how many of the blue cats are bouncing"
This question however comes within the context that they're looking at cats on the screen. Unless they get zero cats as their answer for the very first question they play, then the context tells them "cats shown here".
None of the blue cats are shown here, so we don't have any information to answer a question about them.
If we were previously shown 5 blue cats and none were bouncing, then "5" could seem like the valid answer. If you assume that there is no memory element to the game (which is a weird assumption when teaching), then it's only because there is no option to select "there's not enough info" that "zero" can contextually become the right answer.
I guess I'm still a little bit salty about being asked these kinds of imprecise questions on high-stakes standardized tests that are supposedly testing you on your own precision and accuracy. I remember running into questions that had more than one valid answer because of imprecise wording, at which point the test becomes "what did the test writer intend" instead of "what did the test writer ask."