The interesting question is always are the rewards (of whatever action) worth the costs. There is no merit in always choosing the 'good' side, if that is always rewarded and never punished.
You could steer people into a character arc by shifting the costs. Say the game involves fighting zombies. Early on, zombies are, well, zombies; all you can reasonably do is kill them. Kill zombies, rescue your wife, she starts helping kill zombies. Then things get more complicated: the zombies start getting smarter, and your wife starts wondering whether killing them is wrong. Every time you kill a zombie, she gets more and more angsty about it; if you just keep going, she leaves you, and soon you run into a boss zombie that you can't survive without her help. If you switch to killing zombies only when they're actually attacking, she sticks with you, you survive until the Turning Point.
At the Turning Point, you're fighting a bunch of zombies led by a much more intelligent zombie, who acts pretty much human...and turns out to be your daughter. If you kill her, your wife leaves you and becomes a powerful enemy; if you let your daughter go, she comes back to kill you. You have to find some third way, some way to keep your daughter alive (undead) while you look for a way to cure zombiehood. At this point, if you kill any zombie, there are serious costs—maybe you start tearing up, thinking of someone killing your daughter, and can't defend yourself.
Oh, and, once you've found the cure, then you have to decide what to do about people who want to remain zombies. Do they have the right to? Do you have the right to cure them against their will?
It's not the same as recognizing when a character is creating their own arc; but, really, that's an AI-complete problem. First you have to recognize good and evil; then you have to recognize a shift from one to the other; then you have to recognize when that shift is reflecting a coherent story. Oh, and then you have to figure out how to sell a video game that requires the player to make up a coherent story.
Dynamic cost model
You could steer people into a character arc by shifting the costs. Say the game involves fighting zombies. Early on, zombies are, well, zombies; all you can reasonably do is kill them. Kill zombies, rescue your wife, she starts helping kill zombies. Then things get more complicated: the zombies start getting smarter, and your wife starts wondering whether killing them is wrong. Every time you kill a zombie, she gets more and more angsty about it; if you just keep going, she leaves you, and soon you run into a boss zombie that you can't survive without her help. If you switch to killing zombies only when they're actually attacking, she sticks with you, you survive until the Turning Point.
At the Turning Point, you're fighting a bunch of zombies led by a much more intelligent zombie, who acts pretty much human...and turns out to be your daughter. If you kill her, your wife leaves you and becomes a powerful enemy; if you let your daughter go, she comes back to kill you. You have to find some third way, some way to keep your daughter alive (undead) while you look for a way to cure zombiehood. At this point, if you kill any zombie, there are serious costs—maybe you start tearing up, thinking of someone killing your daughter, and can't defend yourself.
Oh, and, once you've found the cure, then you have to decide what to do about people who want to remain zombies. Do they have the right to? Do you have the right to cure them against their will?
It's not the same as recognizing when a character is creating their own arc; but, really, that's an AI-complete problem. First you have to recognize good and evil; then you have to recognize a shift from one to the other; then you have to recognize when that shift is reflecting a coherent story. Oh, and then you have to figure out how to sell a video game that requires the player to make up a coherent story.