Log in

No account? Create an account
Bar-sur-Loup Provence France

Austin GDC 1: Chris Crawford on "14 Conceptual Shifts: moving from games to interactive storytelling

Crawford did a presentation -- more a polemic -- on the future of storytelling for games. Several of his fundamental tenets would be echoed on Wednesday by Andy Stern, as both of them are part of the movement that insists that proper game story must be non-linear and dynamic, and therefore machine generated. Whether you agree with his ideas or not, Crawford certainly was passionate and intelligent in developing his fourteen points:
1. Character comes first.
No disagreement there; in writing for any medium the characters' motivations define what happens and why. Conflict between characters provides stories with tension and drama, and any story full of weak characters will lack intensity and emotional engagement.
2. Primacy of interactivity
As the medium for which we are writing is games, which by definition include gameplay, the player is constantly making decisions and taking action. If the player is not active it mimics a passive experience like film or literature, and a story without interaction would probably better be told in another form.  A person buys a game with the idea that they will be actively involved in what happens, and it follows that the story must take this into account. This seems entirely logical.
3. Screw graphics.
One of the points upon which the presenter and I are in agreement. Graphics are a means to an end, though they are most often treated as an end in themselves.  However, storytelling does not require them. They have become an enormous roadblock for budgets, development cycles, and even for story. If 3D graphics cannot present what is happening in a game story, due to either technical or budgetary limitations, the most probable decision in a game project is to change the story rather than simplify or remove the graphics. Few studios have the bravery to put in Max Payne- or Thief-like 2D cutscenes to compensate for story elements that are difficult to present in the game or cutscene engine.
4. Ditch plot
His words, not mine. Crawford presented Story as something that is a noun and data and static, while storytelling is a verb and a process and is dynamic. It is the difference between a cookie and cooking; one is a thing, the other is an action and therefore interactive and therefore more attuned to what a game player wants. His thesis is that storytelling is the goal, not story. The process is what we as game writers should be focused on, rather than on the results of the process which are the plot and the events that make up the story. This is one of the points where his views and that of most writers diverge; on the one hand is the belief presented which is that the process of creating credible events is more important than the outcome of those processes and the shape that it has. This seems to beggar the most fundamental question that kept coming back to me: Who cares how good the storytelling machine and mechanisms are if the resulting story is uninteresting? And how do we program a computer to discern between an interesting and a boring narrative?
5. What does the user DO?
Because games are interactive the emphasis should be on what the player is doing; on what his or her actions are and what effects they have. This is what is important in a game story, and not what the player sees or hears. The player should be creating the story as he plays, rather than having the story presented to him as something that he is walking through. Again, this makes sense for a story that is told in a game.
6. What are the verbs?
The verbs, that is the range of possible actions that the player can have, are the linchpin of the gameplay and of the richness of the player experience in the game. The verbs of a game -- as of any software application -- define what the application is. More verbs in a game mean more expressive power.
7. Linguistic User Interface.
Here Crawford touched on the ideas of the user interface, starting with the old-fashioned and cryptic yet powerful command line interface of 20 or so verbs with many parameters. The combination of these commands created enormous possibilities for actions, however the interface is arcane and highly specialized. Next came GUI's, whose graphical menus and context boxes presented somewhere around 100 or so verbs for the average user. At this point Crawford made the very cogent point that even in a beginner's reading text -- one of the classic Golden Books -- there are already 122 verbs. If the verb capacity of a GUI is only on the level with a first grade text for storytelling, the average user interface is not going to provide us with a very meaningful experience.
The ideal would be a Linguistic User Interface, where a player uses everyday language to interface with the computer and can take advantage of the thousands of verbs in the English language.
The bad news came next; a computer system can never provide a natural language interface. The problem is that a language mirrors reality, and it has been mathematically proven that for a computer to provide a natural language interface the computer must be able to encompass reality.
Crawford's way out is to say that because stories are artificially created toy realities, all that is needed to "use" these realities is to have a created toy language. A toy language in a toy reality can fit inside a computer; our reality can't.
8. Language = reality
The caveat is that the toy language needs to be designed at the same time as the toy reality in order to be useful. In order to do this, Crawford stated that "we" -- definition never provided -- use a powerful authoring system called Deikto.
It was at this point that my skeptometer started pinging, and it went off fairly steadily for most of the rest of his session. Much of the remainder of his presentation discussed the system that he had developed in order to tell stories in his way on a computer. Deikto was joined by Sappho, a programming language for storytellers, and his talk concluded with a sales pitch for storytron.com, the site that he and his partner have created in order to promulgate their world view of what game story should be. However, in the interest of being unlike Fox News and presenting a fair and balanced picture, I will continue with his remaining 5 points.
9. Inverse Parser
For a natural language interface to work the computer must be able to read a string of incoming text, break it down into verb - subject - object etc., and then act on it. The human is the supplicant; the computer is the judge. There is in addition the paradox of the "parser puzzle", that the possible combination of words in a sentence approaches what is a good approximation of infinity for any practical desktop application, and that it is therefore almost impossible to parse natural language effectively. The user is therefore constrained to adapt to the needs of the interface, using a limited number of words in a limited number of ways. This is not good storytelling.
Crawford's solution is inverse parsing; limiting the scope of what the computer has to analyze by using dramatic and grammatical context filters. In the sentence "Look out, it's a ___", the blank will not be filled with the adjective "chartreuse," the verb "climb," or the number "twenty-three." In addition, if it is a horror survival game full of zombies, though the blank is a noun it will probably not be the nouns "belt sander" or "foie gras." Relying on grammatical and dramatic context, the computer can more rapidly translate the player's actions and desires.
10. Ditch space.
The presenter stated that spatial relationships are an annoying waste of effort and computing power; stories are about social, not spatial, cognition. I agree that spatial cognition has little to do with storytelling. However in a graphical game world, the distance that two NPCs keep between each other, or from the dog that may or may not be friendly, or from one group of NPC's or another, may speak volumes about the NPC and the situation and the story.
11. Programmers are not storytellers.
True. Crawford goes on to say that, therefore, storytellers must learn to program. I think that any writer who wishes to tell stories in a game setting is probably wise to accept this one.
12. Algorithms animate.
Algorithms are what programs use to control the choices of in-game characters. Context-dependent choices require the mastery of context-dependent algorithms; therefore, for Crawford, it is necessary that writers learn math and the use of algorithms.
One could argue this; one could also argue that all it requires is a half-decent authoring tool. I would dare to guess that Storytron handles the decision-making process of in-game characters through the use of algorithms with a bad or nonexistent front-end tool. This may, however, simply be my natural cynicism shining through.
13. A programming language for storytellers.
Apparently it exists, it is called Storytron, it will change the way we tell stories in games, it is inevitable, it is good, and it will prevail.
14. Kinder, Gentler Math
Storytelling requires special arithmetic, and storytellers must learn it.


The second part of the talk was given by Laura Mixon, an SF author who is the writer half of the team that is developing Storytron. She apparently believes quite strongly in the system, so I am willing to give her the benefit of the doubt. However, all of the examples that were given, by Mixon or Crawford, were what I would call "small". Conflict between two characters that leads up to a murder, emerging through the algorithms of Storytron, is a great and unexpected thing. I would not call it a story, however; I would call it an incident or a piece of a story. Nothing that I saw during these three days convinced me that computers will be able to create plot and character arcs -- the things that keep consumers hooked on a game, movie, or TV series -- in any sort of meaningful or interesting way.
What I liked very much about Storytron, and what I would like to try, is to use these kinds of algorithmic characters to add flavor and sub-plots to a given scene or location. The idea that autonomously-generated behavior can be going on around the player while he tries to to what he wants to do seems fascinating. No two players would live a game experience in the same way, and unexpected and unpredictable actions would make a game environment seem much more life-like. I will therefore download and test Storytron, and add further comments at some later date. For the moment, however, I cannot imagine that this is a story generating tool, even if it is a storytelling tool. And if it is "only" a storytelling tool, and generates interesting incidents and events, its place in the scheme of game development may have a limited, albeit useful, future.


Thanks for that. I feel quite educated from your write-up... almost like I was there.
Bar-sur-Loup Provence France

December 2011

Powered by LiveJournal.com