Consider a pretty common scenario: players meet a monster and want to manipulate it into doing something. They have a spell or ability and are trying to use it in an unusual or counterintuitive way to achieve a result the DM does not expect. When does it do what the player wants it to?
A few things I keep in mind for this decision:
1. It's fun and rewarding for player to have agency and get a result they want
2. Players make sense of things differently than the DM might make sense of things
3. DM's job is also to create an appropriately challenging experience. Denying a result will make an experience that feels more challenging, whether fairly or not
4. A world that works consistently and in a way that matches our reality will reward lateral and creative thinking, thus creating a more engaging game
5. In fantasy, there are physics that have no analogue to real life, and cannot be modeled through common sense.
6. In real life, sometimes the world behaves in ways that don't make sense to us, because we don't see the whole picture. This is true in video games as well -- we hope that our new spell or gun will work in some way, or that physics will interact together in a particular way, but in fact they do not.
7. Sometimes the player simply has a bad idea
8. Sometimes, the DM might have an unconscious or difficult to articulate reason for why something shouldn't work the way a player wants. In my experience I've sometimes found that reasons might lie in barely perceived interactions with other game mechanics. Sometimes there are other reasons.
9. The DM might want to provide a certain aesthetic experience that resists something working the way the player wants it to.
10. When the DM doesn't allow the world to act in the way the player wants, the player may feel as though the world is intractable or unresponsive, or that the DM is being hard-headed or stubborn instead of reasonable, particular if the reason doesn't make sense to the player
11. A world that feels intractable to the player is probably the worst possible outcome. The greatest strength of role playing games is complete freedom of imagination, so a break-down in that aspect is really undesirable. As game narrows in tractability, it becomes less and less engaging, until player will disengage and do the next most interesting thing to them, or if nothing is available, find a way to cope with the situation.
12. A tool that can be used here is the relationship between player and DM. If DM can communicate that they realize that the outcome may be disappointing, or any other variation of what is going on for them in the moment which maintains the emotional connection, this can really ameliorate some of the bad outcomes.
With that in mind, I tend away from automatic "yes" and maybe hew a little too close to automatic "no." It's been helpful for me to experiment with playing games that are really built around collaboration and lateral problem solving, like White Hack. In 5e D&D, players are so powerful that I tend toward no. I wonder if this could be improved on my part?? On the other hand, see point 8 above.
Makes me wanna play more White Hack.