July 11, 2003
Surfing around I came across a new book to be published this Fall, called “Tomorrow’s Stories: How Responsive Narratives Will Change Storytelling”. The author is Andrew Glassner, perhaps best known as the creator and editor of the Graphics Gems book series on computer graphics. According to his bio, he’s also written a novel, directed a short animated film, and designed some interactive fiction / game prototypes. I heard Glassner opine on the “Future of Fiction” panel Noah co-organized at SIGGRAPH 2000, and remember him as thoughtful and articulate.
His forthcoming book “addresses the fundamentals of how and why successful games and stories work. It is the first book to give a reader a practical understanding of both the structure of story and the structure of participatory gaming. This knowledge helps the reader see why and how today’s models of interactive fiction succeed and fail, and provides a foundation for developing new storytelling art forms that harmoniously integrate interaction and narrative.”
He sees limitations in branching narrative / hypertext approaches. “These ideas, and their cousins, have been tried time and again in the marketplace but have yet to achieve mainstream success. There are some very good, specific reasons for this lack of success, and those reasons can be found by going back to the basics of what stories and games are, and how they work.”
I look forward to reading the book, and applaud anyone that articulates their understanding of how to navigate, as Glassner puts it, “the dangerous waters of interactive fiction”. That said — and I don’t know to what extent this is the case in Glassner’s book, since I haven’t read it — but I’m wary of approaches to tackling interactive fiction that rely primarily on design, without offering new technology to support the design. Don’t get me wrong, design is a critical component of any interactive experience — and in fact is what is sorely lacking from most technology-heavy academic CS research in interactive fiction. But design alone is inadequate.
My feeling after spending 4 years with Michael on the problem of interactive fiction, looking for a magic design solution to the problem, is that sadly there is no magic design solution. It is fundamentally a very complex thing, that will require powerful new generative architectures to make real progress beyond where we are today. In Facade, a design-heavy and technology-heavy project, we’ve been chipping away at the complexity of the problem, with bits of success, but realizing the field has a long, long way to go.
Design-heavy / technology-light solutions essentially back off from the complexity of the problem, by necessarily reducing or virtually eliminating player agency. They do their best to immerse players in a story, using tried-and-true storytelling techniques (“the basics of what stories are”). In fact some designs specifically focus on giving players the illusion of agency – leading them down a single path, while trying to hide the fact that there is only a single path. This can be an enjoyable experience, but mostly for the same reasons that a good movie or book is an enjoyable experience.
July 12th, 2003 at 11:54 pm
Boy, I can certainly agree with your comments on how difficult it is to build real interactive narrative stuff. One of the problems holding back the field is the widespread feeling that good interactive storytelling will require no more effort than a computer game requires. It turns out that a great deal more effort than that will be required.
I too am looking forward to this book, especially because I have begun work on my own book on interactive storytelling.
July 13th, 2003 at 2:39 pm
More effort than a staff of 20 – 30 (artists, programmers, writers, composers, management) working for two or three years and spending a seven figure budget? What kind of effort do you have in mind?
July 13th, 2003 at 6:09 pm
Responding to Chris D.’s question to Chris C. … Naturally, this depends on what you mean by “interactive story” — a term almost as open-ended as “computer game”. I used to think that no one had yet built a “real” interactive story; I’ve come to realize that I have a certain set of criteria for what “interactive story” means, that not everyone else shares :-) , and that of course plenty of things have been built to date that can be called interactive stories.
(As a point of reference, Michael and I briefly list our design criteria for our current project here. To our knowledge, nothing has been built to date that satisfy even these minimal criteria.)
The bigger vision some of us have for what a interactive story would be, would require a much larger database of content — dialog, behavior, character animation, etc. etc. — than what we see in today’s games. Today’s games have relatively huge amounts of content compared to older games, but it seems that an decent interactive story will require even more.
It’s probably not possible to build that much content in a brute-force way. We’re going to have to make systems that have some ability to generate dialog and behavior, in collaboration with human authors. My guess is that without that kind of assistance from the system, the bigger vision is probably not achievable.
July 15th, 2003 at 12:53 am
OK, read your criteria and it sounds doable, provided you start with good natural language interpretation and an AI that can pass a general Turing test. To get there from the current state of the art will, I fear, require more than time and money. It will require a major breakthrough. Because it seems to me that your goal with this program requires the program to genuinely understand the player’s input, rather then just create the illusion of understanding (like, say, ELIZA) or understand only in a very rigid Blocks-world like context (like expert systems).
I think you’re right that you’ll never brute force this. Take your example where the player becomes uncomfortable and asks for a drink to change the subject. Now imagine the number of other ways the player might try to change the subject – He might ask for a cigarette, comment on some object in the environment, mention the weather, mention some current event (there’s a whole other bag of worms), talk about his medical problems or any number of things. You could spend months trying to anticipate the player’s response for this single beat. Now, the standard approach to this is to have some sort of stock reply to non-revelent input but really this is not good enough. No matter how many variations on the stock phrase you use it ends up sounding like “I am not programmed to respond in that area”.
What really interests me about your ideas is your approach to the protagonist. In drama or fiction a story happens to a set of characters who behave, ultimately, as they must given who they are and our experience is of coming to understand the characters through seeing what they do and say in a series of events designed to display those very qualities. What you seem to be getting at is injecting the player, in his proper person (i.e. not “Role-playing” as someone else) into a story with other, program controlled, people and the experience is to find out how YOU would respond to a given series of events designed to get at some aspect or aspects of qualities you may possess. This seems like an important difference to me, a genuinely new thing arising organically out of the nature of the medium. Very interesting.
Alas, I suspect we are in the position of Charles Babbage in respect to this sort of fiction, able to see how it could and should work, but lacking the basic technology to make it happen.
July 15th, 2003 at 10:38 am
A question / challenge we’ve put to ourselves is, once you open the can of worms of offering players open-ended language input in interactive stories, can you build artistically interesting experiences that don’t require you to solve the Turing test?
Our answer to that question is yes, by developing some better-than-Eliza natural language processing and decision making technology, and a story design that works within the still severe limitations of that technology. As stated before, we’d love to see future research on more generative architectures, to reduce those limitations from “severe” to “less severe”.
July 15th, 2003 at 12:02 pm
makes me wonder if latent semantic analysis could be useful in the context of interactive narrative to help meet andrew’s and michael’s criteria…