Affect actions, not commands
We don't want situations like this:
> attack johnwhere whether we are allowed to attack John or not depends on whether we have behaved violently earlier. Even if I'm not a violent person (by nature, by inclination, by habit) I can still decide to attack John, and this kind of response rings false. But we should not forget that the interactive fiction author has to interpret the commands given to the game, and that she can base this interpretation on the character traits of the protagonist, rather than basing the list of allowed commands on these traits. What I'm getting at is this. Suppose the protagonist has been behaving violently earlier in the game:
No, you are not violent enough.
> attack John.Suppose, however, that the protagonist has not been behaving violently earlier in the game.
You smash his nose and send him sprawling on the floor. "If you ever touch my girlfriend again, I'm gonna kill you!", you scream.
> attack JohnWhat happens here is that the very same command is interpreted in different ways--as a command to do physical violence in the first case, as a command to do verbal violence in the second.
"Seriously, John, I thought you were my friend. But now I see that you are nothing but another lying, hypocritical ass who takes advantage of his fellows as soon as he thinks that they're not looking. Have you no shame at all, kissing my girlfriend in my fucking house? I don't ever want to see your face again, John. Now beat it."
The possible advantage of this is that it allows for dynamic characterisation. The protagonist really takes on the character traits that the player puts into him during the course of play. This might be cool. I'm not sure whether it will be, but it might be; I would like to see it tried.
Now one could say--rightly--that it should be possible for someone who hasn't done physical violence before to start doing physical violence now, and that a game like this would artificially bar him from doing so. True. But--this kind of limitation is inherent in all interactive fiction all the time. It is a feature of the medium. You can always only take those actions that have been provided for you, and this does not feel artificial, precisely because it is the essence of IF. (It is the same way, say, that it doesn't feel artificial that people in opera's sing all the time.)
One potential problem with this approach is that all happens behind the scenes. The player will not notice it on a first play-through, unless you find a way to draw her attention to it. So you'd better make sure that you're okay with that, and that you want your game only to reveal its possibilities on subsequent playthroughs. (Hidden cause-effect structures are always tricky in IF; they tend to go unnoticed and not affect the player at all.)
Affect NPCs, not PCs
An interesting suggestion made by Jimmy Maher is to take the approach criticised and then apply it to NPCs. So basically, the NPCs would keep doing what you had learned them to do earlier. Jimmy suggests a child-rearing scenario, but there are other possible applications: training your combat team before you take them into the Afghanistan mountains, for instance.
There certainly is artistic potential here. What about this: there is an NPC around who looks up to you and has a tendency to copy your behaviour. This would make you a role model, and that lends an entire new dimension of morality to your actions. (You are the cynical policeman who has neither family nor friends, doesn't care about his own survival, and likes to take on the gangsters in a very dangerous shoot-first-ask-questions-later mode. Now you've got this young recruit in tow, who has a two-year old child and her whole life before her. Are you even willing to show her the effectiveness of "your" way when you know it might lead her to do the same thing and get killed?)
Things would get even more interesting if the NPCs can, at a certain point, decide to rebel against what you have taught them--now there is a game I want to see made.
Tracking how the protagonist is perceived
Another good idea, due to Aaron A. Reed, is to use character traits not to limit the behaviour of your character, but to track how the world views him. So if you have been raping and pillaging your way through the land, people will run away from you as soon as they see you and bar their doors. If you have been kind to people, others will confide in you.
This is, of course, simply one way of confronting you with the effects of your actions. As such, the author of a piece will have to make a careful assessment: is it better to implement an abstract system of character traits, or is it better to track the outcomes of specific events? Do you want people to fear you because you ahve taken violent actions, or because you have slain Ralph the Merry in a bar fight? I take it that the former option becomes more useful as the game becomes larger and more episodic, while the latter is more useful for games that are shorter and more coherent.
The difficulty of growth
Tom Hudson remarks that changing is difficult; there really is such a thing as an ingrained habit--and an author might want to reflect this. True--but how is this best done? I am not convinced that the original proposal (making certain actions unavailable) is the way to go. Emily Short has a system in Metamorphoses, if I read it right, whereby some actions become more difficult if you have shied away from that kind of solution earlier; she does this by having the game suggest that you might not want to do that, but allow you to do it anyway if you try again. If it were well-clued that you can do it if you really want (which is not always the case in Metamorphoses), this might work. It highlight for a moment the fact that it's an action you hesitate to perform.
However, Emily does point out that people just don't notice that which actions they are allowed to perform depends on what they've done before. I suspect that you need to ensure that player understand this; for why else are you implementing it? This might be done in a thematic way: for instance, if the protagonist has access to a performance-enhancing but very addictive drug, it makes perfect sense that there is a lot of hesitation when you try to do it for the first time, and no hesitation at all when you it for the tenth time.
Taradino C. suggests, and I like this suggestion, that the difficulty of growth can be shown without modelling it. How? Because when the protagonist falls into a habit, so does the player. That's a really interesting thought. If you fought your way past the first five guards, you'll probably attack the sixth one as well, without thinking about it. Games might be based around this concept.
(Indeed, these kinds of patterns of behaviour even persist beyond the games where they are learned. Here is a fragment of an actual review of The Baron:
[...] I didn’t know how else to approach the problems that came in front of me. For example, with the wolf, I just figured I had to kill it. Yet, the detail of everything was grueling and frightful. I did not enjoy it. I wish for the game just to have said, wolf killed, and to be able to move on with the story. The gargoyle was not so much trouble, however, I did not listen to his story, and as I sit here now I wonder if I would have heard some solution to speaking and negotiating with the Baron in order to get my daughter back. Yet, I thought I need to press on and keep going. Then, I met the baron and I thought it was my duty to kill him. [...]A game where the player has to break out of her habits--now that's an interesting design aim.)
I am sure there are more possibilities, so let's discuss.