RE: Just to spur some discussion
Delighted to see people posting here again. I had something relevant to post last week, but didn't get around to it, and seem to have temporarily forgotten what it was... John wrote: What I still grapple with is how to ensure that all possible outcomes lead up to an equally satisfying story-telling experience... A possible answer just occurred to me, but it might require extremely good AI: Have you folks ever played a game called "Once Upon a Time"? It's a fairytale storytelling game that uses a set of printed cards to help push the players/tellers into staying within the genre. (Sorry if I've described it before here; I don't remember...) For instance, every player is given a different "Happily Ever After" ending card, each printed with a common ending to a fairy tale. The goal of the cards-based part of the game is to move the story in such a way that it results naturally in the ending that you hold; if you get rid of all your other cards (printed with types of characters, events, places, etc commonly found in fairy tales), and then finish the story with your ending card, you win the game. Of course, this game is best when played more to create an enjoyable story than to win, but the goals aren't incompatible. At pretty much any point during a game of OUaT, I have a fairly clear idea of what I would do (if it were my turn and if I weren't interrupted) to use up the rest of my cards and get to my ending. That idea of how the rest of the story "should" go (in order for me to win) changes over time, of course. So what I'm suggesting is that the computer moderator of a nonlinear story be able to generate paths-to-a-satisfying-ending from any given state; if the state changes, the end-paths also change. Yes, this does require a lot of AI. A primitive version of the idea, though, might still be useful, modeled after OUaT: the computer knows a certain set (preferably a fairly large set) of satisfying endings, and it has a certain set of tools (characters, situations, etc) to use to bring about those endings. At each step of the story, it determines which ending is closest (or picks one randomly, perhaps weighted in some way), then determines what needs to happen to produce that ending, and takes small steps in that direction. (This is the covering-up-the-deus-ex-machinae approach that Len mentioned; in this case you cover up what you're doing by doing it in small increments, steering slightly toward a particular ending.) Let the player(s) act some, then recalculate what ending you're aiming for, and thus what you need to do to get there. It seems to me this approach is along somewhat similar lines to what Erasmatazz(sp?) does already; the engine might need a little more sophistication to behave as described above, but I think it would be feasible as long as ending-space is fairly thickly populated. The hard part is when you get very close to an ending and the player suddenly veers and does something totally off the wall. The sudden about-face can damage the narrative warp engines, causing you to have to limp to the nearest ending on impulse power for repairs. --jed, suddenly struck by a silly mood (thump) PS: Maureen, I think you missed some of the conversation; your old address was still subscribed to the list, so a couple messages bounced. I took the liberty of subscribing your new address (as given in bounce msg); hope you don't mind.
RE: Just to spur some discussion
Some more thoughts follow. I'd like to chart out all possible ways of creating a nonlinear story... if you see a recipe that I've left out here, please go ahead and add it. 1 - The most primitive way of doing nonlinearity is to create N storylines and choose between them via a random number generator. Easy development time, but is not really nonlinear. Also is nondeterministic; if I encounter a "cool thing" it's not guaranteed that the next time I play, and repeat all of my actions exactly, I will achieve the same result. Does anyone remember the old game "Dark Tower" or something like that? It was a board game, with a large black plastic tower in the middle. It was swivel-able so that it would face only one person at a time. Whenever someone moved, a button was pressed and the tower would reveal an outcome (e.g. a brigade approached you to battle). I have fond memories of that game, even though the tower was nothing but a glorified die :) 2 - The next-easiest way of implementing a story is to provide the N storylines, but create a tree structure where everytime the user chooses an action, a different path along the tree is chosen. This provides determinism. This is very similar to the Choose-Your Own-Adventure books, as well as the vast majority of Infocom games. There were exceptions; I think Deadline was time-based instead of turn-based. 3 - Next approach seems to me to take control out of the user's hands, while not delegating the choice to a random number function. The only place to put this control, then, is in the characters and/or environment. This involves infusing some level of AI into the characters. Again, this can be done via random functions; if the dice is less than 3, have the character welcome you; otherwise have the character shoo you away. This is "simple" AI and is not very intelligent. 4 - You can expound on that and provide a rule-based AI; for instance if a character likes what you've said so far, it will do a certain thing. Notice the parallels; 1 is similar to 3 and 2 is similar to 4. In fact we're not really changing anything, we're simply moving the control of the story from "The Code" running the program, to "The Code" personified in a character or environment. 5 - The most exciting (to me, anyway) approach is to take advantage of things you really can't predict. Implement neural networks and/or genetic algorithms. A neural network is a program that can train itself over time. It's conceivable to place a NN in each character and train them through the beta phase; by the time you release, the characters will be fully trained to do what you want, but still not be completely predictable. The other approach is to embed them with traits (i.e. genes). Create a whole lot of them and have them reproduce (copy the traits via the usual cellular reproduction paradigm). Kill off the characters that don't do what you want, and eventually you will have a fleet of characters that do what you, as the author, intended. The extreme danger in #5 is that (a) you don't know if the training will work to your satisfaction, and (b) if the end result will be what you intended. It's entirely possible to create a big brute that acts like a kittycat. It seems to be that the more nonlinearity we build into a system, the harder it is to predict. And to me, I think the prediction of a storytelling system is of the utmost importance to an author. Therefore, is there any point to nonlinear storytelling? -John On Thu, 19 Nov 1998, Jed Hartman wrote: Delighted to see people posting here again. I had something relevant to post last week, but didn't get around to it, and seem to have temporarily forgotten what it was... [snip]
Re: Just to spur some discussion
Hi, I saw this too, and must admit it got me thinking - however, in the ST universe, it seems as if every person is very intelligent and imaginative from the very start. What a great presumption for a storyteller! If assume your audience has some intelligence, you can leave all kinds of avenues open. Unfortunately, most of what passes for entertainment these days assumes I have the IQ of a pork chop. Other thoughts - my God, what an insane amount of algorithm you'd have to build in - AI for the characters to react "in character" to any new situation, physics models that take into account EVERY law of physics, from Gravity to subatomic interaction, Plus - Cartoon physics - if stories can be assumed to have any sort of fantasy level to them (one of the character's finger/branches burned with a bright flame, yet all he did was whimper, not cry out in pain). If we can deal with the monumental task of creating such a thing, how very cool it would be - Touchrealistic (as opposed to photorealistic) Avatars with unlimited high-level AI who reacted in unanticipated manners, yet within the parameters of expected personality. Star Trek seems so very unmaterialistic, but imagine the market for such a "virtual friend" or slave or partner, what have you. I wonder what moral distinction would be drawn by Star Trek crew members between their holographic doctor and one of the holographic characters from that story. Hey - there's an interesting storyline right there! I'll be sending off my script to UPN any day now. Nice to get a discussion going. -Alan On Wed, 18 Nov 1998, John D. DeCuir wrote: Since I love this list and no one has posted anything in a while, here's a random topic thrown out. Did anyone catch last week's Star Trek: Voyager? The storyline involved a little girl who, among other things, likes to hang out in a holodeck. The story is a child-like story involving two characters, one symbolizing water and the other trees. There's also a Fire-demon as well. (Sorry, I forget all names). The interesting thing is that almost all of the crew reminisced about the same story in THEIR childhoods. The kicker is that everyone had different memories, because everyone encouraged the same story to go in a slightly different direction. (Janeway: When I did that story, so-and-so happened... Another crewmember: Really? Not for me...) etc... This seems to me like the ultimate goal of nonlinear storytelling. What I still grapple with is how to ensure that all possible outcomes lead up to an equally satisfying story-telling experience... Comments? -John
