Hmm. Often times, an open ended question results in a much more accurate answer. Humans tend to approximate or even lie to force an answer to fit a limited choice question. Example: you want to know if water is falling out the sky. You ask: Is it raining? Response is no. You walk out the door and are rendered unconscious by golf-ball size hail.

Better to have asked, "what's it like outside?" or, "what's the weather right now?"

I think his response meant in situations where you just need to figure out which project management tool they use so you can log in for the first time, binary questions are the right tool for the job (information gathering). But I mostly agree that, in the big picture, the broad questions are more helpful for solving the complex problem you’re facing.

and nobody is going to say it’s not raining when it’s hailing outside… but I understand if that was just a metaphor. You’re probably asking that question to figure out what kind of clothes to wear, and they would probably infer that. It’s so contextual though.

Again, I think that depends on where you are in the exploration/confirmation spectrum. I don't really think about it when doing it, but usually I use open ended questions for stuff I don't know, and yes/no questions to link things to my existing mental model.

That also depends on the person. With some people open ended questions are great, with others yes/no are better.

I'll also add that yes/no questions are often faster if you get the response you expect.

Maybe appending "... Or something else?" to a limited choices question can avoid the problem GP @indymike mentioned, whilst still generally be quick

Is it raining or something else outdoors?

Do you like Vi or Emacs or something else?

Do you want to hang out for a coffee or something else?

Usually I don't think it's necessary. My yes/no questions are something like "Okay, so we need to put some more info into that file since that's what exported to the billing database, is that correct?". When I don't know something, I try yes/no questions at first. If I get "yes", then it's resolved. If I get a no, I will ask open-ended questions to explore the subect. Once I think I have a correct mental model ready, I will ask yes/no questions to validate it. If I get a no, then it's the cycle again.

For the "Is it raining outside", I honestly don't know anyone that would just tell me "no" when there's hail outside. For the "Do you like Vi or Emacs or something else?", it would be "What text editor/IDE do you use? Why?". "Do you like Vi or Emacs" by itself is not a yes/no question, it's a "what box are you in" question. For "Do you want to hang out for a coffee or something else?", that would be something like "Do you want to hang out at <TIME>?".

An important part for me is that in yes/no questions, no are supposed to be the exception. In my mind, when I ask them, it's not a 50/50 chance. It's more of a 80/20. If it was a 50/50 chance, that means my mental model is not precise enough.