that story stuff again (longish)
Matthew J. MacKenzie
has come up here before but I'd like to hear the latest thoughts:
One reason people share stories to help make sense of their
experiences, especially ones involving other people, and take in
patterns to help predict events (a basic reason for brains).
It's related to the reason we form stereotypes (which are
condensations of stories), say, or teach our culture (collected
observations about what happens and what works) to the next
generation.
Aristotle's idea of a good story is based on a dramatic arc
which lifts the audience to an emotional climax, then resolves it
(catharsis) -- with the possibility of smaller arcs under it,
also resolved by the end. Now a lot of good stories have been
written which don't quite follow this model. Aristotle's idea of
a bad dramatist was Euripides, with his sudden turns in plot, and
a lot of people kind of like the effect. But the idea of a
dramatic arc is still the best expanation we've got for why a
story is satisfying.
If we want our machines to tell stories, they can't be
machine stories -- they have to be described, first, by
storytellers with experiences to share, and the points of view
they have extracted from them. But the question is this: given
that we partly surrender the direction of the plot to the user,
who has an incomplete clue about the storyteller's frame of mind,
what kind of story do we tell? Can we replace Aristotle's model?
Do we have to try? Whatever you might say in a response to this
message I'm asking you to answer the following two questions:
o What kind of interactive experience would you consider most
satisfying? (Take this one broadly; you could talk about a
single "run" through an interactive story, or the story as some
set of possible runs. What interface you use is less important.)
o What sort of formal system could be used by an IF author to
describe this story, or set of possibilities, or world, or
whatever your experience is, to a computer?
From the papers I've read, the Oz people at Carnegie Mellon
are coming at it starting with Aristotle, creating closure with a
series of emotional events to be injected into the story. These
are arranged as a DAG, with events events as nodes which must
occur in a given order to form an "arc." So several arcs can be
active at once.
A useful variation, I think, would be to keep the search
for arcs but divide the graph into two elements: effects, and
events. The event "a burglar enters through the window," for
instance, would have several effects: it introduces a new
character, creates tension, and probably interrupts whatever is
already going on. A story arc which needed any of these could
pick this event, as long as no other active arc required that its
effects not happen. If it was used for tension, the next event
triggered could be that he pulls a knife. The dramatic arc would
be something like suggest a threat; build tension by directly
threatening the user; release tension by preventing the threat
from being carried out (sirens & flashing lights introduced).
If the burglar event was used to introduce a character, he
could take an interest in the trophy collection on the top of the
book case. One kind of comic relief arc could have a tense
moment in it as a subgoal, combining these events.
This is a very minor variation on the Oz idea, and FAR less
developed -- I'd need a few weeks to even decide if it could be
implemented. It's more fragmented in an effort to try to be able
to come up with new directions no matter what the user does
(that is, you can never "leave the graph").
But the real question is allowing an author freedom to
express a point of view, and while any description of a world
will do that to some degree it's not explicit enough --
controllable enough -- for my tastes. We need to capture _more_
than an author's idea of how a burglar might act, or when the
cops generally show up (always late?) and what they do. The
question is how to capture, for each moment in the story, what a
certain IF author would LIKE to have happen next, what would put
the user into a good story, how to connect the events to capture
that certain idea which the author has developed watching people
on the New York subway for the last eighteen years, or that
childhood encounter with the refugee yodeler bag lady. We could
allow preference to certain plot combinations using weights but
that's just a start.
Whaddaya think? What's a story? (Please try & answer those
earlier two questions first.)
_____________________________________________________________________________
Points made in this message are distributed by weight, not volume.
Disagreements may have been settled during the transmission process.
ma...@eng.umd.edu Matthew MacKenzie
--
_____________________________________________________________________________
Points made in this message are distributed by weight, not volume.
Disagreements may have been settled during the transmission process.
ma...@eng.umd.edu Matthew MacKenzie