The GM decides for NPCs, the Players decide for PCs. It isn't the GM determining how the situation will go if he is playing fairly, because how the situation goes isn't the point, the point is an honest character interaction. The players are still free to do what they want (they could stab the NPC to death and take what they need from them for example). But the point is the GM should have a clear idea going in of what the NPCs motives and and not fit those motives to some kind of outcome the GM wants. At least for me, that is how I approach it. It still isn't going to appeal to you if you don't like this kind of play but much of the reason I play this way is, as a GM I hate having preset outcomes and I hate guiding the players along. All I care about when I am running the NPC is what that NPC is motivated by, cares about, and what responsibilities they have. If the players make a proposal to that NPC or try to trick that NPC in some way (like in the goodfellas example), I am going to honest think through how the NPC reacts based on what I know about their motives (and if I can't decide for some reason, I might roll a die, but usually I have a good sense of it).
I am not saying you can't have a more narrative approach. I am not trying to persuade someone who wants something different than I do in an RPG, that they should want what I want. I am saying this is what works for me. Even when I play narrative games, I much prefer something closer to what I am describing like Hillfolk that tends to be about things like what the characters want (and there are mechanics underlying some aspects of that but I find them fairly unobtrusive to the above style that I outlined). This is actually a game I really like because it does a good job of captured the 'all we did was role-play' aspect of play that can be fun, but because it is oriented around scenes and scene framing it doesn't just meandering and become characters talking endlessly in an inn: it leads to dramatic places.
Sure, if you don't like GM authority in these situations, by all means use these systems. I don't have a problem with GM authority extending to NPCs motivations and behaviors. If a don't like how a particular GM does things, I might not play in games where that person GMs, but mostly I game with people who, even if they think about NPCs differently than I do, are good faith when it comes to this stuff.
I would love a game where that is possible (maybe there is one that achieves it I just don't know about). What I have found in practice is, for whatever reason, people are much more okay with social interactions being handled through talking and negotiating without a system or without random elements, but with fighting they really seem to want random and mechanical elements. I'd love to play a game where the GM decides based on either what is plausible or what works dramatically (not both mixed together but two possible 'legal schools of thought' for the GM to follow).
Not so much about winning, but about engaging in the game. If we're playing a game where one of the PCs is Sherlock Holmes, I think it's safe to assume that there's a mystery afoot. So engaging with that mystery is a big part of play. How the character does so is vital. The portrayal of Holmes's other traits... the emotional stuff, the social awkwardness.... they are most important in how they impact his ability to do what he does. Having rules for this...penalties of some kind, or at least deficiencies in stats absolutely can help portray character.
Sure! Again, as I said above to @Crimson Longinus this tangent of the conversation came about when someone said that social mechanics should not be used. None of us are saying that the method we'd classify as traditional isn't valid. We're just saying why there are other methods and what makes them valid and/or appealing.
I would love a game where that is possible (maybe there is one that achieves it I just don't know about). What I have found in practice is, for whatever reason, people are much more okay with social interactions being handled through talking and negotiating without a system or without random elements, but with fighting they really seem to want random and mechanical elements.
One observation as a possible reason here is we all have experience with social interaction just by virtue of being human. That is how we do things. But we don't all have experience with combat (and even if we do have experience fighting, we might only have experience with one type of fighting: someone may have been in brawls but never been military combat for example, or been in military combat but never a sword fight).
Right, but then you are largely creating the criteria that is allowed to work. Talking will only work if they find out about the guard's vice, but they can always stab him!
I don't agree that this is true, but even if it was, it really doesn't matter. If people feel that that the freeform social interaction is very close to the real one then that's good enough, even if it objectively wouldn't be terribly realistic.
You're narrowing the paths forward. And this isn't necessarily a bad thing, it just is a thing. It has an impact on play. It shifts it more toward "let's figure out the solution the GM has in mind for this" and away from "let's come up with a solution for this".
Holmes is a good example for a lot of reasons. I have used him to help address this in some of my own writing in games. I think this gets at a fundamental preference divide people don't even realize they have until it comes up in play as a problem or in discussions like this. There are different ways this can be expressed, but something like "Do you want to be Sherlock Holmes or do you want to be in a Sherlock Holmes story". I think the former is more in line with the old 'kind of playing a version of yourself but Sherlock holmes' where you are the one solving the mystery (you as a player have to find the clues, piece them together, solve the mystery). Whereas if you want to be in a Sherlock Holmes story, you expect that if you are sherlock holmes, then you have all his problem solving abilities and intellect. You the player might not, maybe you couldn't' solve a mystery to save your life, but the system should allow you to play a character who can. Again, my particular phrasing (be sherlock versus be in a story) is somewhat inelegant. But both these approaches are totally valid. And if you give someone A when they want B, it is going to create issues. So understanding this aspect is useful. I run a lot of mystery and investigation adventures so this is just something that crops up naturally when you do it a lot.
Outside of social areas you're familiar with, I'd argue it has, if anything, a worse one. Its just one people think they understand better than they do.