I have always liked the idea of competing heroes in a game. Fables acted like that would happen. If you don't take a quest, others will. The quest isn't just going to wait on you. Then the other heroes could die, succeed and prosper, etc. There was a game that did this a little bit, but I can't remember the name of it. It had different parties going out to do the same quests as you. But was more of a Diablo II clone, and I didn't like it all that much.
But the idea would be a Skyrim game with a world that is progressing and changing whether or not you're involved. Some other hero will get the credit. If you spend your time going after the thieves guild quests, someone else is going after the Mage guild questline. Etc.
If you haven't played them previously (and if they're your type of game), you might want to check out the Stalker franchise:
"...for the first in the franchise 'Shadow of Chernobyl' it was stated that the AI had to be impeded because it could have completed the game by itself, characters were non-scripted and could develop even when not in contact with the player..."
- from
http://all-things-linux.blogspot.co.uk/2012/09/stalker-clear-sky-late-verdict.htmlIt sounds like a lot of what's being discussed in this thread revolves around the idea of improved artificial intelligence and artificial life in games. Some games have made good use of this (such as Stalker and, obviously, DF to some extent!), but the problem is that you run into the same irreducible complexity idea. Yes, you can have a lot of runtime variables being modified by the player's actions (and NPC actions), but at the end of the day the player has to see the outcomes of these actions or they're not going to 'get' the point of the game.
At which point, you have two options:
1 - write a large number of detailed, in depth responses to a large number of possible 'game states' (i.e. "if player has been this horrible to this many people that this NPC knows, then say something like this. If player has been horrible to slightly fewer people that this NPC knows, but is wearing something that looks amusing, then say this" etc.) This requires a large amount of unique content creation - which is not something that can currently be done automatically without feeling incredibly stilted and artificial.
2 - write fewer or less detailed responses. This runs the risk of feeling repetitive or, again, stilted and artificial, but less content generation makes it more achievable.
3 - write a detailed railroaded storyline, and make the sandbox outside that storyline slightly less detailed. This is the case for games like Freelancer, for example, which had (what I thought was) a very well-presented story, but if you wandered off into the sandbox for too long the repetitive nature of umpteen different people saying to you, "We don't own this station, but we have an understanding with the people who do" would drive you space-crazy.
4 - make it online-only and hire a human GM.
Is it possible to do a highly branching game, to the extent that the player won't realise that they're branches rather than pure sandbox? Yes, if you're willing to do a lot of content generation (and I don't mean the typical 'fancy graphics and voice acting' content here, either, I mean actual gameplay content, written dialogue, anticipated reactions, et cetera - you know, all that stuff a good GM has to plan for in a tabletop session).
Is it possible to do such a game without putting in a lot of content generation? I don't think so - to even get a program to start generating content smart and subtle enough to make a player think it's hand-written, you'd need to write such a complex program that you might as well just write the content in the first place!
Still, even achieving a good version of options 1 or 3 in the above list would probably elevate the game above most of the 'mainstream' RPGs out there at present. DF actually seems to be progressing (slowly) in that direction, which is fantastic.