IF and AI
Paul Christopher Workman
This question has nagged me: it's been pretty much accepted
that better AI will produce better IF, and that's probably
true; but if all you do is make your IF's characters as
realistic as possible, using extremely effective AI techniques,
and ignore all other parts of the IF, what do you have?
A MUD, except that all characters other than the player is
artificial.
Am I wrong? I suspect that the solution to IF, the technique
that will make it work as effectively (as a literary form) as
traditional fiction, will involve a new interpretation of
literary elements that take interactivity into account.
AI techniques in character formation would be just icing
on the cake.
What do you think? What am I missing?
--paul
Jamieson Norrish
Workman <pw...@andrew.cmu.edu> writes:
This question has nagged me: it's been pretty much accepted
that better AI will produce better IF, and that's probably
true; but if all you do is make your IF's characters as
realistic as possible, using extremely effective AI techniques,
and ignore all other parts of the IF, what do you have?
A MUD, except that all characters other than the player is
artificial.
If you ignore all other parts, then you are going to end up with
something that makes MUDs look less primitive than they are. However,
this is true if any other single aspect was concentrated on to the
exclusion of all other aspects. I think it is valid to say that the
most sophisticated interactin comes with the IF characters, rather
than with objects. Therefore, in order to increase the interactive
nature, improvements in the characters are necessary.
Am I wrong? I suspect that the solution to IF, the technique
that will make it work as effectively (as a literary form) as
traditional fiction, will involve a new interpretation of
literary elements that take interactivity into account.
I think this depends very much on what you want from IF. What literary
elements were you specifically thinking of?
AI techniques in character formation would be just icing
on the cake.
I think rather that they are an effective means to an end; with
specific reference to the moral/immoral discussion, I believe this
should be handled exclusively through the reactions of the other
characters. After all, who creates moral codes? People do, and so it
should be the people in the story who add the moral overtones. In
order to do this, you need more advanced characters.
Jamie
U52...@uicvm.uic.edu
some feedback to your post.
>but if all you do is make your IF's characters as
>realistic as possible, using extremely effective AI techniques,
>and ignore all other parts of the IF, what do you have?
>A MUD, except that all characters other than the player is
>artificial.
I think your ideas are right on target. While autonomous AI agents, interacting
with a player and each other, can contribute to IF, I don't feel that they
alone are the answer. Agents setting and working to achieve goals can make
a nice simulation, but not a focused work of fiction.
>Am I wrong? I suspect that the solution to IF, the technique
>that will make it work as effectively (as a literary form) as
>traditional fiction, will involve a new interpretation of
>literary elements that take interactivity into account.
I think a truly successful IF system would require some sort of master
storyteller, a meta-agent, controlling and manipulating the course of events
in order to meet dramatic goals.
Karl Steiner <ste...@bert.eecs.uic.edu>
Paul Christopher Workman
Jamieson Norrish@kauri.v
> something that makes MUDs look less primitive than they are. However,
> this is true if any other single aspect was concentrated on to the
> exclusion of all other aspects.
True.
> I think it is valid to say that the
> most sophisticated interactin comes with the IF characters, rather
> than with objects. Therefore, in order to increase the interactive
> nature, improvements in the characters are necessary.
I'd have to disagree with this. Although interaction in a purely
physical way is with characters, in a story, the ultimate interaction
is with the form of story itself, or the narrative. A perfectly
valid traditional short story may have only one character: the
narrator. But even in this most basic form, the story changes --
develops -- as it progresses. In traditional fiction, this
development is the same for a particular work, regardless of
who is reading it or when (basically). In IF, the development
of the story is variable and probably unpredictable.
> Am I wrong? I suspect that the solution to IF, the technique
> that will make it work as effectively (as a literary form) as
> traditional fiction, will involve a new interpretation of
> literary elements that take interactivity into account.
>
> I think this depends very much on what you want from IF. What literary
> elements were you specifically thinking of?
In general, plot, and the form of the story as it is effected by
plot. When I (try to) write traditional fiction, I like to include
elements like foreshadowing and ironic plot twists. I don't see how
I could do that in IF. (Well, you can do it by eliminating enough
interactivity to be able to refer to specific plot points at
other points (the most common way of doing this seems to be to
force the player into a set sequence of problems to solve, such
that a player will be forced to work with a certain problem eventually,
or will have definitely worked with a certain problem in the game
earlier), but I don't like this solution as it seem to me to be
avoiding the fundamental problem of interactivity.)
Interesting thought experiment: how would you write an interactive
"Twilight Zone" episode?
> AI techniques in character formation would be just icing
> on the cake.
>
> I think rather that they are an effective means to an end; with
> specific reference to the moral/immoral discussion, I believe this
> should be handled exclusively through the reactions of the other
> characters. After all, who creates moral codes? People do, and so it
> should be the people in the story who add the moral overtones. In
> order to do this, you need more advanced characters.
This is interesting, in that it could allow the characters
in effect to write the story, as opposed to the author (directly).
This is obviously closer to reality, but I'm not sure it's like
fiction as much. How would you deal with a lone-just-character
type plot, such as "High Noon"? (Not the best example but
all I can think of right now.)
--paul
Jamieson Norrish
Workman <pw...@andrew.cmu.edu> writes:
I'd have to disagree with this. Although interaction in a purely
physical way is with characters, in a story, the ultimate interaction
is with the form of story itself, or the narrative. A perfectly
valid traditional short story may have only one character: the
narrator. But even in this most basic form, the story changes --
develops -- as it progresses. In traditional fiction, this
development is the same for a particular work, regardless of
who is reading it or when (basically). In IF, the development
of the story is variable and probably unpredictable.
In a game with one character only, surely the plot all that needs to
be done is the modelling of the interactions between the player and
the objects, and between objects and other objects? That is, the plot
simply results from what the character does, and the author doesn't
need to worry about anything more than the interactions with objects?
Or am I missing something - it's been a long time since I've played
any IF.
> I think rather that they are an effective means to an end; with
> specific reference to the moral/immoral discussion, I believe this
> should be handled exclusively through the reactions of the other
> characters. After all, who creates moral codes? People do, and so it
> should be the people in the story who add the moral overtones. In
> order to do this, you need more advanced characters.
This is interesting, in that it could allow the characters
in effect to write the story, as opposed to the author (directly).
This is obviously closer to reality, but I'm not sure it's like
fiction as much. How would you deal with a lone-just-character
type plot, such as "High Noon"? (Not the best example but
all I can think of right now.)
The unfortunate part about the lone character plots in fiction that I
can think of is that they rely almost exclusively on the author
providing the thoughts of the character (correct me if I'm wrong
here!). As soon as interactivity is introduced, the author cannot put
thoughts into the players head, except in an indirect way, by
producing "facts" which the player will sense, and then interpret for
herself. I'm not at all sure how a plot could be constructed beyond
this, which (I think) simply involves the author in creating the
relationships and interactions between various inanimate objects.
I feel sure that I'm missing something here...
Jamie
Paul Christopher Workman
Jamieson Norrish@kauri.v
> be done is the modelling of the interactions between the player and
> the objects, and between objects and other objects? That is, the plot
> simply results from what the character does, and the author doesn't
> need to worry about anything more than the interactions with objects?
Well, yes and no. I guess what I'm trying to say is that
in traditional fiction there's more to plot than just a sequence
of events, but in IF, that's all you can have, unless you remove
enough interactivity until it (plot-wise) turns into traditional
fiction. Therefore, we need a plot-paradigm in IF that allows
for plot to be more than just a sequence of events.
In traditional fiction, a narrative has a certain flow to it.
It's hard for me to describe it -- it's as if the narrative
has a sort of life of its own. I haven't seen any of this
life in any IF I've used. As an example, consider Roald Dahl's
short stories -- the horror ones, anyway ("horror" for lack of a better
word). The stories start in an innocent way and seem to
progress in a more or less naturalistic way, until at some
point the reader has a relevation about where the narrative is
going; at this point the reader (an observer) feels a sense of
horror with the progression of the narrative, but knows the
horror in the narrative couldn't really be resolved or avoided
with a plot device that's true to the plot-life of the narrative.
The protagonist is not just trapped between two plot points
in which something horrible will get him if some puzzle isn't solved,
that is, it's not just a sequence of events in which you could
possibly append a certain event and resolve the horror.
Consider also the idea of "deus ex machina": "god from a machine".
This refers to suddenly resolving a conflict or problem by the
intervention the gods (it goes back to classical greek dramatic theory).
The point is that although you *could* just haul in a diety
to tidy up a story, you don't want to -- it would feel like cheating
and would be unsatisfying to the reader.
Compare this to (a hypothetical) IF. If a player is faced with
immenent doom, s/he could just wave a magic wand and be returned
home. Or, not, in fact it could be that failing to acquire the wand
would be part of failing "to solve the puzzle." But the point is,
either way, other than the wand being seen as being too easy, there's
nothing about this example which could be seen as somehow violating the
gestalt of the narrative.
In a typical IF, you can describe the plot, when you're finished
with the game, like this: "first I found the lantern, then I
went in the cave, then I fought the ogre...", or if the writer
has removed enough interactivity, it could be described from the
outset: "first you have to find the lantern, then you have to
enter the cave, then you have to fight the ogre...". Again, there's
more to a narrative than just a sequence of events, more even than a
sequence of events plus colorful descriptions. A traditional
(non-IF) swords-&-sorcery novel with a narrative like that would
seem ... lifeless. The narrative should be somehow alive; I don't
know of a good way to do that in IF yet. I don't think it's
possible to have literary art in which plot is just a side-
effect of the characters' interactions; I don't think that
the traditional plot paradigm can be applied to IF; I think
that IF is a viable art form; thus, I think that there must be
a new paradigm for plot in IF.
>The unfortunate part about the lone character plots in fiction that I
>can think of is that they rely almost exclusively on the author
>providing the thoughts of the character (correct me if I'm wrong
>here!). As soon as interactivity is introduced, the author cannot put
>thoughts into the players head, except in an indirect way, by
>producing "facts" which the player will sense, and then interpret for
>herself. I'm not at all sure how a plot could be constructed beyond
>this, which (I think) simply involves the author in creating the
>relationships and interactions between various inanimate objects.
Well, I think if we include inanimate objects, we come very
close to the original case in which a player interacts with
other characters (within the context of this discussion, anyway).
I think it's possible to have a worthwhile plot in which the
protagonist doesn't even interact with inanimate objects. In
a way, you could say he would then be interacting with himself,
or with his memories. Thus the reliance on the character's
thoughts, as you mentioned. I realize that this would be extremely
difficult to do in a medium like IF in which the form is traditionally
expressed in second-person ("you see...", "you pick up..." etc.).
The idea I was trying to approach is that at the core of a story,
after you discard character-character interaction and still have
a story, you have a progression of narrative which exists beyond
a simple sequence of events denoted by interactions.
--paul
Jamieson Norrish
that won't cause too many problems.
I think that maybe what you mean by the plot being more than
individual event elements, is that a fictional (non-IF) plot has a
structure on a higher level than these individual elements. That is,
the whole, the collection of separate events, has a form and a meaning
of its own. Out of this whole, fiction gains such elements as theme;
the meaning or interpretation of events as grouped together. (okay,
okay, I'm making this up as I go along, with no knowledge of the
subject whatsoever. :)
This, then, is what IF is missing, since the author cannot (or rather
should not) dictate events to such a degree that they conform to an
overall meaning and coherence. With many different possibilities, the
author cannot possibly hope to have every combination meaningful on a
"meta-level". So, as you said, a departure from the traditional view
of plot is needed. Unfortunately, I don't have any idea as to how to
find this new perspective, with which it is possible to give meaning
to every grouping of possible events.
BTW, does anyone know what the Greek for deus ex machina is? I've
always wondered...
Jamie
Paul Christopher Workman
>individual event elements, is that a fictional (non-IF) plot has a
>structure on a higher level than these individual elements.
>This, then, is what IF is missing, since the author cannot (or rather
>should not) dictate events to such a degree that they conform to an
>overall meaning and coherence.
Yes. This is exactly it.
Did Brenda Laurel cover this at all in her thesis? I haven't
read it yet.
>BTW, does anyone know what the Greek for deus ex machina is? I've
>always wondered...
Uh, doesn't "deus ex machina" mean "god from a machine" ?? (although
"machine" would have a different sense to it that what we're used
to here in the post-industrial-revolution world). Did I quote it
wrong or something?
--paul
Jamieson Norrish
(Paul Christopher Workman) writes:
Did Brenda Laurel cover this at all in her thesis? I haven't
read it yet.
I don't know; I haven't read it either.
Uh, doesn't "deus ex machina" mean "god from a machine" ?? (although
"machine" would have a different sense to it that what we're used
to here in the post-industrial-revolution world). Did I quote it
wrong or something?
Oh yes, "deus ex machina" does mean god from the machine, and refers
to the theatrical device of a crane-like arrangement which was used to
lift actors up into the air, representing going up into the heavens.
The modern meaning of the term is derived from that. Unfortunately,
the term "deus ex machina" is Latin, and the original idea was Greek.
Therefore, it should have a Greek name, as well as the Latin one. If I
remember I'll find out next week, unless someone else knows.
Jamie
Erik Max Francis
> >BTW, does anyone know what the Greek for deus ex machina is? I've
> >always wondered...
>
> Uh, doesn't "deus ex machina" mean "god from a machine" ?? (although
> "machine" would have a different sense to it that what we're used
> to here in the post-industrial-revolution world). Did I quote it
> wrong or something?
_Deus ex machina_ means "god from machine" literally. It is meant to
mean a process in a story whereby the solution comes from outside the
story itself; it doesn't necessarily have anything to do with a machine.
The original reference, as I've learned, comes from Greek plays, where
the characters would get into a mess, and a god or gods would come down
from heaven and fix things -- the resolution, you see, coming from
outside of the characters themselves. The reference to "machine" comes
from the fact that the gods were represented by wheel-and-pully systems
which would lower an icon of a god into the scene, you'd hear a voice
offstage say "Everything is better now," and the god-icon would be
reeled back up. That was the extent of the machine.
I believe that's a fairly accurate (if simplified a little)
representation. If anyone knows something more specific, I'd be glad to
hear it.
Erik Max Francis, &tSftDotIotE ...!apple!uuwest!max m...@west.darkside.com __
USMail: 1070 Oakmont Dr. #1 San Jose, CA 95117 ICBM: 37 20 N 121 53 W / \
If you like strategic games of interstellar conquest, ask about UNIVERSE! \__/
-)(- Omnia quia sunt, lumina sunt. All things that are, are lights. -)(-
Jonathan R. Ferro
>BTW, does anyone know what the Greek for deus ex machina is? I've
>always wondered...
We've had several explanations of what it means, but the point is that
the words "deus ex machina" are clearly Latin, whereas the original
concept is Greek. So, assuming the Greeks actually had a name for what
they were doing, what was it in their language?
-- Jon Ferro Einsprachigkeit ist heilbar
Karl Steiner
>This, then, is what IF is missing, since the author cannot (or rather
>should not) dictate events to such a degree that they conform to an
>overall meaning and coherence.
>...
>Yes. This is exactly it.
I disagree that IF should not try to manage some structure or
coherence. While I haven't seen any, I don't feel that it is
impossible to have an IF that provides a large degree of user freedom
yet still maintains thematic integrity, structure and coherence. There
are two interactive experiences that I can think of that provide
structure in a manner close to the type I am imagining:
Improvisational Theatre and Human Moderated Role Playing (as opposed
to computer games).
In the improvs I've seen, the actors have or are given characters and
a scenario, and possibly a loose plot to follow. While the actors are
constantly introducing new situations and plot elements, these are
picked up or dropped by fellow actors based on their anticipated
humor, conformance to plotline, personal whim or whatever-- but this
process has the effect of shaping and refining meaning and coherence
and generally keeping the proceedings close to the theme and shape
set forth at the outset. On occasion I have seen performances where
the shape has not been defined, or where the actors have deviated
significantly from the original premise, and these were less
satisfying.
Similarly, in most of the role playing games I have observed or
participated in, the game master is responsible for shaping the
course of the interactions. While the players are free to choose
whatever course of action they like, even straying outside of the
originally envisioned scenario (or scenarios), a skilled game master
must still manipulate events into a satisfying structure for
participants. Again, games where the players move chaotically from
encounter to encounter, hacking and slashing with no building drama or
coherence have been the least enjoyable for the participants. Rather,
games that have some structure, however loose, have met with more
player approval.
I am not advocating a dictatorial structure for IF. When I want an
experience where the author is in complete control, I read flat text.
What I am suggesting is that it is possible and perhaps desirable to
have highly interactive situations that still conform to aflexible
structure, meaning or coherence, and that these situations can often be more
satisfying than unstructured simulations.
Karl Steiner <ste...@bert.eecs.uic.edu>
Jamieson Norrish
What I am suggesting is that it is possible and perhaps desirable
to have highly interactive situations that still conform to
a flexible structure, meaning or coherence, and that these
situations can often be more satisfying than unstructured
simulations.
This is all well and good, and I think that structure and coherence in
plot is a good thing, when it still allows for player freedom.
However, just how are you suggesting that this be done? In the two
examples you gave, there were other people present who shaped the flow
of events. In the Theatresports example, the players introduce their
own events and play through them in such a way as to make them
meaningful both in the context of the plot and of the other events
they have done. With roleplaying, the person in charge of the game can
tweak events so that they gain meaning. However, this is not done
before the players start the game, it is done while the players are
playing. I doubt that computers can yet be programmed to be
sufficiently creative to change the events they present in order to
produce a coherent plot structure and theme, and authors can't
typically mess with the game while the players are playing it.
Jamie
Paul Christopher Workman
Karl Ste...@bert.eecs.u
> coherence. While I haven't seen any, I don't feel that it is
> impossible to have an IF that provides a large degree of user freedom
> yet still maintains thematic integrity, structure and coherence.
About theme: suppose you try to write a tragic IF. But the user
keeps behaving as if the IF were a comedy.
Do you force the situations to be tragic, and if so, doesn't that
violate interactivity, or do you allow the theme to follow
the user's actions, and if so, how do you maintain thematic
integrity?
If the first case, how can you maintain a theme if the user doesn't
follow it (it won't be a very good tragedy if the user makes
adolescent jokes all the way through), that is, is it even possible
to force theme?
--paul
Karl Steiner
>About theme: suppose you try to write a tragic IF. But the user
>keeps behaving as if the IF were a comedy.
>follow it (it won't be a very good tragedy if the user makes
>adolescent jokes all the way through), that is, is it even possible
>to force theme?
am imagining are cooperative storytellers, rather than coercive. As
such, the quality of the experience depends in part on the input of
the participants. As long as this is understood, I think users would
not feel overly restricted.
The user would participate in theme selection. If the user wants to
make adolescent jokes, a tragedy may not be the best choice. If the
user did choose a tragedy, the user shouldn't be surprised at
thematically inappropriate results (maybe that's what they want). Or
the user could choose (or the system could switch to) a tragicomic IF
where humor is a more integral part of the scenario.
Excerpts from Jamieson Norrish: >IF and AI
>This is all well and good, and I think that structure and coherence
>in plot is a good thing, when it still allows for player freedom.
>However, just how are you suggesting that this be done?
>sufficiently creative to change the events they present in order to
How can a computer guide story development while still providing
high interactivity? Good question. Although it may be difficult, I feel
that it is a worthy goal to pursue.
Karl <ste...@bert.eecs.uic.edu>
Peter Weyhrauch
experience which provides structure:
In the improvs I've seen, the actors have or are given characters and
a scenario, and possibly a loose plot to follow. While the actors are
constantly introducing new situations and plot elements, these are
picked up or dropped by fellow actors based on their anticipated
humor, conformance to plotline, personal whim or whatever-- but this
process has the effect of shaping and refining meaning and coherence
and generally keeping the proceedings close to the theme and shape
set forth at the outset. On occasion I have seen performances where
the shape has not been defined, or where the actors have deviated
significantly from the original premise, and these were less
satisfying.
This is very interesting to me, since I've heard it said suggested
that long term "shape" would be hard to do for improv actors, that
these actors must live in the moment and therefore cannot do much
comprehension of how each event that happens fits into a larger
framework.
Were these comedy improvs, or somehow dramatic improvs? What let you
know there was a defined shape? How long were the improvs?
I guess I am curious to know about any groups that are doing longer
term, dramatic improv.
Thanks,
Peter Weyhrauch
Project Oz
Karl Steiner
where the audience suggested characters and situations. There was a wide
variety of situations, and probably differing skill levels of the actors, but
in general the audience responded more enthusiastically (laughed more and
harder, more applause) at sketches that maintained some touch with the
original premise and achieved some sort of closure.
By shape, I mean a story that has a setting, conflict and resolution. Some
sketches never established a conflict in the given setting and with the given
characters. Some started with one conflict but kept introduced more and more
subplots so that in the end no conflicts were resolved and the cohesion
between events was tenuous. Other started with an original conflict,
introduced perhaps one or two subplots, then at least wrapped up the original
conflict at the end. While I don't know the though processes of the actors, I
would suspect some internal process must have been taking place that guided
the actors in choosing which leads to pick up on and which to drop, and how
and when to wrap things up. It was my observation that the sketches that had
some shape were better appreciated than those without.
Karl
Karl Steiner
find debatable? Or do you think structure of any type, even an evolving,
flexible structure, is undesirable in an IF?
Karl
Jamieson Norrish
ste...@bert.eecs.uic.edu (Karl Steiner) writes:
Well, the above might not be addressed specifically to me, but I'll
answer anyway. I have no objection to structure in IF if it is
evolving and flexible. I would object to a rigid structure imposed
without consideration on the game.
My only question is how such a flexible and evolving structure could
be added to IF. It's not something I have the faintest idea of how to
do, either elegantly or not, but I imagine it would be a marvellous
breakthrough. :)
Jamie
Paul Christopher Workman
Karl Ste...@bert.eecs.u
> would suspect some internal process must have been taking place that guided
> the actors in choosing which leads to pick up on and which to drop, and how
> and when to wrap things up. It was my observation that the sketches that
> had some shape were better appreciated than those without.
Yes. This is a good example, and if I may extend it...it looks as
though the behavior of the actors resembles the (musical) behavior
of jazz musicians. Although I'm not a jazz musician, as I understand
it, while playing some standard segment of a piece of music, they discover
some sort of mutually-understood although perhaps not directly described
musical structure. At this point, an individual musician is able to leave
the current standard piece, and improvize, following the understood
structure. The musicians can, more or less, take turns improvising
under this structure while the others, also understanding the structure,
can back them up.
The problem is, the musical structure idea does not map directly
to plot. (If it did, what are you improvising, plot-wise? The
individual plot elements are what we are discussing doing improvization
on.). Or you could look at improvisational IF this way: instead
of having a standard plot (being the sum of each plot elements
and traditionally formed by tweaking the IF's setting to make sure
each action must follow another), have an amorphous, general plot.
Each action in the IF must conform to this general plot. This
would then mean that we're passing the plot problem of IF to
the other components, i.e., individual actions.
This brings us, partially, back to "IF and AI." If you could
use AI to model characters very well in IF, instead of modelling
the characters directly, you would, in effect, want to model
*actors* playing the characters, actors who understand plot,
instead.
Thing is, I wonder how primitive you could keep the AI part,
if you could sufficiently developed the understanding-the-
plot part.
Agree or disagree?
--paul