process vs. data

[WFreitag@aol.com]

If Chris' insistence that one can only interact with process is meant to 
imply that data-oriented approaches to interactive storytelling cannot be 
valid, I strongly disagree.

My exhibit goes back to what we learn in the first class of Interactive 
Storytelling 101: branching. Consider a basic paragraph tree with branching 
links of only the most straightforward type.

Now, as we know, it is not _practical_ to create such a tree for a story of 
any length. But I've never heard it argued before that such a work, if 
created, would not be interactive. Suppose an author is willing to do a lot 
of work, and generate only a short story. Allow three options at each choice, 
and make every branch valid, with no deaths or backfolding. With 1093 
paragraphs, in one pass through the story, the reader reads seven paragraphs 
and makes six choices. 

Why is this not interactive? And why is it not an interactive story? The 
"process" in this case is trivial. All the meaning is in the "dead bits" of 
the text. Yet for the duration of seven paragraphs and six choices, the 
quality of the interactivity is comparable to that of any other interactive 
story system I've seen (picking from limited choices isn't ideal, but there's 
not much better out there right now anyway), and the quality of the narrative 
(if the author is skilled) is likely to be far better. 

I agree with the general idea that you can only interact with systems that 
"think." But is there any reason to insist that some or all of the thinking 
can't be done in advance? Dead bits cannot think, but they can, miraculously 
enough, encompass human thought. That's a pretty cool capability, to which we 
owe most of our civilization. 

In fact, I submit that when it comes to telling stories via non-human agents, 
some of the thinking MUST be done in advance, because human thought is the 
only way we know to create a good story. (I monitor the AI research, and when 
and if this changes I'll know, but it's true now.) Therefore the only way at 
present to bring narrative into an artificial storytelling system is through 
_data_. Of course, data's not very good at being interactive. But that's 
tough. That's the way it is.

Like Chris, I perceive a distinction between data-oriented and 
process-oriented systems. At the data-oriented end we have the "take a story 
and try to make it interactive" approaches. They suck, because data isn't 
good at being interactive, though it's VERY good at telling stories (e.g. 
movies). At the process-oriented end we have the "take a simulator and try to 
make it tell a good story" approaches. They suck, because process isn't good 
at telling stories, though it's VERY good at being interactive (e.g. computer 
games). 

And the real problem is there's no continuum between the two extremes. 
Process and narrative data don't get along. When you try to make them meet in 
the middle, process says to data, "I can confer my interactive powers on you, 
but only if you get rid of all that complicated human thought you're 
carrying." And data says to process, "I can infuse my narrative genius into 
you, but only if you do exactly what I'm expecting you to do." Then the two 
get into a fist fight, which for unknown reasons causes all the money in the 
vicinity to disappear.

We may have to work the problem from both ends to get anywhere. It is 
counterproductive to say that only process can ever be interactive, just as 
it is counterproductive to say that only data can ever tell stories. We have 
to hope that one (or both) of those assertions is not true, and we'd better 
not rule out any possibilities too soon.

In fact, we already know that neither is true as stated in such extreme 
terms. We see data being interactive in branching stories and hypermedia 
(though the interactivity isn't very good). We see process creating stories 
in simulators and AI systems (though the stories aren't very good). 
Unfortunately, the gap between doing something and doing it well enough to be 
worthwhile can be enormous. Consider how much easier it is to create a 
computer program that correctly makes random chess moves in accordance with 
the rules of the game ("it plays chess, but it isn't very good") than it is 
to create one that actually plays well enough to be worth playing.

- Walt