Dynamic interactive fiction
Mithradates
There have been a lot of Infocom stories, numerous amateurs' adventures, and
oodles of sites providing MU*s (MUDs, MUSHs, LPMUDs...) I've seen a lot of
them, enjoyed several, especially what's come out of Infocom, but I've been
struck by the invariable assumptions of design that characterize interactive
fiction. These assumptions are not exactly bad, and Infocom did some
excellent work with them (ObInfoLaud), but nobody has tried anything else,
as best as I can tell. These assumptions are that the organization of space
is made with rooms, and that the world's contents and order of events are
fixed or strictly dependent on the player's choices (the son, George, in
Deadline goes for the safe iff you show him some document at the reading of
the will, etc.). Has there been any discussion or work done on alternative
assumptions for interactive fiction?
I have been developing an alternative form for interactive fiction myself.
My reason is that I'd like to have interactive fiction that is more like
roleplaying games, and I don't mean such a banal excuse for a RPG like
'Eye of the Beholder' or 'Mines of Titan'. I'd like a story that resembles
a flesh & bones gaming session; my requirements resemble a Turing test for
GMing. Here is my list of specific gripes against conventional text
adventures:
1. The stories are only new once.
2. The stories can, and are expected to be,
known in advance.
3. The characters are frequently made of cardboard.
4. The worlds have small borders.
I. The first day of playing a new Infocom story will always be warm in
my heart. Such a voyage of discovery to watch the opening events and
explore
the new territory! I will always remember moments like those because they
provided so much pleasure, yet they were so few. But starting a new session
with a favorite GM does not induce that same nostalgia.
II. Many tough (and memorable) puzzles require that the player play through
the puzzle numerous iterations before it can be solved. (The timewarp in
Sorceror, or the Babel fish puzzle in Hitchhiker's.) However, such playing
conditions do not exist in the real world. The penalty for this
characteristic became clear to me from playing The Lurking Horror; maybe I'm
just thick-skinned, but I was intrigued but not frightened by this story
inspired by Lovecraft. Risk must be present to induce fear.
III. Compared to characters in noninteractive fiction, even Infocom's
characters are flat. Would Floyd ever play hide and seek with you? Could
you convince Belboz to travel with you in Spellbreaker? Could you talk with
Monica about her fascination with electronics? The plots are too fixed for
the variation in behavior one would need to develop the characters.
IV. Eventually, whether you're playing AMFV or TinyMUSH, one reaches a
point
after which there is nothing new to discover (at least currently, as on a
MUD.) Must this be a concrete characteristic of i-f software? Or could
software generate new territory and events?
I would call my solution 'dynamic interactive fiction'. I am still working
out the nuts and bolts, so that I might experiment with it on a friend's
MUD,
which is one reason why I am fielding this description to the net. The
basic
principles are thus:
1. Geography and events are generated 'on the fly', from rules.
2. Space is organized in a hierarchy of zones for variable degrees
of detail.
3. The player's time frame varies in it's period of activity.
(this works better on one-player games than MUDs).
I. I believe that if you develop good rules, that start at the highest
level (a world, a continent, a city) that go on to describe their details
(trees, mountains, castles, rooms), and organize them to work off each other
like dynamic equations in chaos science, unpredictable but believable worlds
can be created with reasonable effort. Plots and characters can also be
generated in the same manner. A text generator exists that has produced
output that was accepted by a campus newspaper as a letter to the editor;
the same principles could be used in a routine to describe
computer-generated situations to players.
II. A careful observor of conventional fiction or roleplaying sessions
knows that the region of action will change in its scope as time passes.
Sometimes the author or GM is very vague, and the action will take place
over
a very large and vague geographic area. At others, a conversation will take
place in a small corner of a salon. The distinctions in space made with
rooms
are arbitrary; in dynamic interactive fiction they would also be transitory.
The same pub visited twice could be described amply as one 'room' when you
visit it at an off hour, and again with several when it is busy.
III. Similarly, the player's time frame changes through the session. Some
activity may take place over days (like traveling), and others have periods
of seconds (like combat). A consequence of this design is that different
activities can take place concurrently. It would be possible to talk with
a friend while walking to the corner, or hunt for food while journeying to a
distant castle.
I would like to see these principles implemented in a text adventure or MUD.
My concern now is to decide what nature of information is required from an
author of dynamic interactive fiction. I figure if I can describe a good
language for the specific purpose of writing dynamic interactive fiction,
then
the implementing it will be straight-forward. I am considering that a
relational language like Prolog would be appropriate. I am open to
suggestions on what this language should be like, or how else I should write
a game like I've described. If you know about any games or MUDs that
already behave dynamically, or you have an idea that I have not brought up
here, please reply.
Ben Taylor
"Hello. I am Oedipus. You are my father. Prepare to die."
email: Benjami...@grace.bah.rochester.ny.us
post: 121 Wickins Road; Scottsville, NY 14546
phone: (716)454-4360 (work) (716)889-5723 (home)