Between the DCs for a single check and the DCs for a skill challlenge.
It seems that this one is particularly dissonant for folks; how do you have a system with objective DCs for task resolution and subjective DCs for conflict resolution? I think this is why they see Skill Challenges as opaque, dysfunctional oddities. For them, conflict resolution DCs seem arbitrary, incoherent. They're expecting to simulate a micro-process, with granularity of phenomenon through zoomed-in, high resolution and tight accounting for cause and effect. They seem to want Skill Challenges to be a succession of that procedure.
I really don't know how to convince them otherwise on this specific issue as I know you, myself and others have tried many times. My answer has always been that conflict resolution is (i) zoomed-out, lower resolution and (ii) context-driven. That means that each check/stage/panel/micro-resolution is going to have a lot more vectors than "can I physically jump across this pit?" Further, that consideration will be a 2nd or 3rd order consideration. The 1st order consideration (the reason to break out the conflict resolution framework in the first place) is typically bound up in what is at stake; "if I don't make this jump my loved ones die". Further, the "jump", due to its zoomed-out, lower resolution (with respect to time, space, overall events) will likely involve much more than just the jump over the pit. It may involve dodging incoming artillery from pursuit. It may involve spatial awareness to not get lost. It may involve a forensic understanding of where you are and where you need to go. You're abstracting multiple tasks/tests at once and choosing the most-fitting, broadly applicable arbiter. So there is a lot going on. Further, the inherent assumption is that the GM is framing the PCs into genre-relevant, challenge-relevant situations with respect to their level. Tally all of that up and you have subjective DCs for conflict resolution. It makes sense by my eyes.
I don't see why these are arbitrary. Part of the skill of good GMing in a closed scene resolution system with freeform descriptors feeding into resolution (or loose approximatins to freefrom descriptors, like 4e skills) is narrating complications in a way that isn't arbitrary because it puts pressure on the PCs in ways that are salient to and engaging for the players.
I don't see them as arbitrary either. In fact, they are a fundamental necessity. A necessity for both the reasons mentioned above (due to the abstract, zoomed-out resolution, multiple vectors) and for the agenda of a dynamic narrative. Its basic logic that the more contracted/bound each possible input is (each check/stage/panel), the more constrained, and predictable, the output will be (the narrative evolution, ultimate resolution, denouement). If you want dynamism, open up the variables. If you want constraint, lock down the variables.
In my view the point of skill challenges is to provide a system for resolving non-combat situations that is different from both free-form narration and GM fiat.
Agreed.
My guess is that skill challenges work better when they are resolving an abstract situation.
If you have all the details of the situation nailed down, then it's difficult to change those details based on the result of a skill check. You can't introduce a secret door, a wandering monster, or a frayed rope if you know there are none of those things before you start the challenge. That means you can't introduce those elements if the PC succeeds or fails on their check.
If the situation is abstract - and almost all social conflicts are abstract, since the DM can't possibly know all the details of an NPC's personality - then there's a lot of room for the DM to react to the skill checks of the PCs and prompt the players for more.
The abstract situation allows the DM to directly address the player's reason for playing the game, as telegraphed through their PC's actions and build. If the situation is too detailed to allow the DM to do that, then it fails.
Great post here except I would say that Skill Challenges work
only, not best, when they are resolving an abstract situation. They don't work at all when you're premising your expectations upon granular, zoomed-in process simulation. If you want that, then you should be using task resolution (see pit > check > gauge pit > check > dodge artillery > check > navigate terrain > check > jump pit > check) and objective DCs. I would reserve "they work
better" for when they're resolving a high stakes scenario.