With regards to the current discussions about "idea generation" and "what if" questions, I've written an ontology, or categorization of answer types.
(Those with programming experience will recognize many of these types as basic (or not so basic) programming data types. This is intentional as computational theory has a long history of working with deciding on answers to questions. )
Implicit with all answer types is a requirement to justify the answer. Just saying "Yes" or "No" isn't useful to anyone.
Some answer "data types" that I can come up with:
Boolean: Yes or No. The valid answer must also provide some justification for an answer but fundamentally, the answer is yes or no. reality-check almost always resolve to a [boolean] Yes or No. No one ever worries about these kinds of questions because they are inherently answerable and (generally) quickly answerable.
Integer or Floating Points: Give me a number. These too are easily answerable though perhaps with a bit more work on the part of answerer. Providing a number and justification is easily done. hard-science and science-based questions often ask for this kind of answer.
Multiple Choice: This is a kind of extended boolean. The OP provides a list of options and the answerer provides the best choice. These are also easily answerable.
Event Sequence: An OP may ask a "what happens, then what happens, then what happens, then what happens; given these initial conditions and these world-systems". The answer is a chronological sequence of events. While it's easy for the OP to request a sequence of events over "too-broad" a range, requesting a sequence of events, in and of itself, isn't bad. Note, to remain a valid question, the OP must ask about an event sequence in a system, not that of an individual. Further, the OP must only specify the initial conditions. Specifying events to occur or changes to world-systems some time after the initial conditions constitutes "too broad".
A List: (Ah, the current point of controversy :) ) A list in and of itself, is a valid answer type. An OP asking for a set of systems required to support human life on a space ship, is a valid world-building question (albeit a very simple answer). Contrast this with a question asking for ways that an AI house might kill its owners without any constraints to determine if one answer is better than another. The problem isn't the inherent data structure of the answer, it's that there isn't a good way to choose a best answer. Granted, list answers tend to make people uncomfortable because they are so easily abused but I submit that the relationship is correlative, not causative. List questions don't cause opinion or plot generation answers, they just happen to appear in the same place at the same time.
World-System: Here, the OP states a world system (and associated constraints) and asks about interaction between their system and "real-life" world-systems. (Though it could Phrases like "How would this work?" frequently show up in questions asking for world-systems. A common example of this kind of question is astronomy questions about orbital mechanics. World-system answers often fall prey to [combinatorial explosions] of feedback loops and complexity. Climate or socio-economic-political questions are frequent victims of unmanageable complexity.
(The World-Systems answer type is really really complex and I'm sure can be chopped into lots of different sub answers. I think this is enough for right now.)