Developed any general rules of thumb from your experiences, or is it mostly just one of those “keep jiggling the handle until it works” kind of things.
Like for example outside of IF if you’re designing a room to be a playable environment in a third person game, you scale everything up by around a third. Exterior environment, you scale everything down by around a half (or more). But then you just sorta fiddle around until things look right, and then you fiddle around some more until things play right. So there’s a bunch of just twiddling the knobs until it works, but there are still basic heuristics to get you within walking distance of where you want to be before you start just winging it.
Like I propose one general heuristic above: you have to give the player a prompt for there to be stakes. I think there are some other general rules: I think the Postel Principle applies (originally from network programming, specifically the TCP/IP spec, it says that you should be broad in what you accept and narrow in what you send…in this case I think interfaces for complex things like driving should go out of their way to anticipate variant inputs and how to accept them, but in general the game should pick a single consistent nomenclature for how things are referred to in-game); in general it’s a good idea to implement a whole bunch of details around any problems/interfaces that players are going to have to feel their way through (like learning how you’re expecting them to enter vehicular control commands) but you should also go out of your way to telegraph which bits are greeble and which bits have gameplay significance; in general you want the interface for a complex task to either be “full autopilot” or “full simulationist” unless you need to split the difference for a specific gameplay (or narrative) reason (that is, either make it something that players don’t have to worry about at all or make it part of the basic gameplay loop, don’t annoy players with fiddly “interface bookkeeping” tasks like having to swipe the keycard every time you use the elevator in Planetfall); and so on.
I resist this because in parser IF there is always a prompt, and there is always a space between two prompts. Of course the player has a prompt unless the game has ended. Subdividing activities does not, by its nature, guarantee a high stakes experience for the player. Any activity can be subdivided into multiple prompts: picking things up, walking, talking to NPCs, driving to the grocery store. Actions tend to be infinitely subdivisible.
I think there is an art to making a player feel that they are in control, but that may or may not involve x number of prompts. In fact, it may not even involve giving them control. The pod ride (and the opening of Starcross, too) pushes its luck. With different details or timing, those scenes could have easily flopped, and their experiences of repeatedly typing “wait” could have robbed the player of all sense of agency.
If the player can’t do something interesting with the prompt, I’m not convinced that it contributes to a better play experience. In fact, it can give the impression that there is nothing at stake.
I really don’t see any way Planetfall could’ve raised the stakes of the intro by removing interactivity. Like I said earlier, no player is going to worry about dying in an intro cutscene—either it’s not going to happen (the danger isn’t real) or it’s inevitable (and so there’s no stakes because that’s just how the story goes).
I understand that you’re arguing that it’s possible to provide interactivity and fail to establish any sense of urgency/danger/whatever. Sure. But my point is that it’s impossible to provide those without at least the illusion of interactivity.
In slightly different terms, I’m arguing that giving the player a prompt is necessary for there to be stakes. You’re arguing that giving the player a prompt isn’t sufficient for there to be stakes. Those are not inherently contradictory propositions.
As for what actually works for establishing stakes, that really depends. Clearly Planetfall does so fairly convincingly insofar as the danger is not a fakeout, the player can get a game over.