Re: Higer-order plot abstractions

["B. Fallenstein" <b.fallenstein@gmx.de>]

Hi Walt,

Thanks for your comments!

WFreitag@aol.com wrote:
> Benja, you've brought up quite a lot to process and respond to. But what
> strikes me first as of possible interest is the relation between "higher
> order plot elements" as you've just defined them, and "nested subplots" as
> you and I (and others, but mainly you and I) discussed at some length back on
> the Erasmateers list.

Um, I see it a little differently. You seem to say that higher-order
plot abstractions are one thing, and nested subplots are another. I see
nested subplots as one *plot abstraction*, along with events,
characters, etc. Higher-order plot abstractions would then be any of the
above with an additional quality, namely being higher-order. As I see
it, nested subplots can be higher-order or not, just as any other plot abstraction.

The point is that if you think in terms of templates with slots, some of
the slots can hold units of dramatic meaning, i.e. other plot
abstractions, because I hold this is necessary for archieving
combinatorial explosion for the player. If all your slots are e.g.
numbers, your plot abstraction will only have a very limited number of
possible "feels" to it, and each of them has to be pre-programmed by
you. If you have slots that can hold units of meaning, the overall feel
of an instantiation of the template can be the feeling authored by you,
*as applied to* some "thing" of dramatic meaning with *another* feeling
authored by you.

I understand that in the following, you talk specifically about
higher-order _events_, and their relation to nested subplots.

> It's clear to me that the two are related. Let's look at the basic "X
> threatens to do A if Y does not do B" from the point of view of conflict.
> There are two conflicts here. One conflict is the reason why X wants Y to do
> B. Although its possible to think of counterexamples, in most cases this
> conflict is something focused on X and is not, at the outset, a simple case
> of X versus Y. The other is the conflict created for Y by X's demand; this
> conflict might motivate Y to do B, or to kill X, or to make a counter-threat
> to X, or to go to the police.

Right. Note that in the example, we not only have a higher-order event,
but also a higher-order subplot that is the result of this event-- Y's
conflict isn't something first-order like "X is mean to me, what do I
do?" It is "X is threatening to do A to me if I don't do B." As A and B
are units of dramatic meaning, the whole thing will feel different
depending on A and B. (And yes, I do see a strong connection between
higher-order events and higher-order subplots.)

By the way, my recent thinking is that X's action is focused on
*creating the development*, as X's concern is not immediate, but the
*result* of the development (X hopes that Y will do B). In other words,
while it is fair to say that X's conflict has spawned the conflict
between X and Y, it is also fair to say that initiating the new
deveopment was an intentional action of X. I imagine that a lot of the
time what players will do when playing a dream is intentionally kicking
off developments-- which sounds quite a bit more interesting than just
doing things for their immediate effect.

> The first conflict has spawned the second. The second exists in the context
> of the first. The second might, in turn, go on to spawn others that exist in
> its own more specific context. For example, the next crucial event might be
> "Y goes to an old pal Z, who owes him a favor, and enlists his help in doing
> B" (or killing X, or preventing A).
>
> X's threat spreads or transfers conflict from X to Y. Y's subsequent action
> transfers it to Z. Many dramatic mechanisms do that. Byock's schema from
> _Feud in the Icelandic Saga_ has a whole category of "feduemes of
> transference" that encompasses all such mechanisms occurring in the sagas.
> Not to put too fine a point on it, this is the way plots thicken.

Actually, for me, it is this point that our hypothetical story goes from
being all black-and-white to generating a tad of interest. My guess is
that having just one level of higher-order plot elements acting on a
level of first-order plot elements is not enough; it is only when we
have created a whole webwork of "units of meaning" referencing each
other that we get something interesting. Your point seems to be similar.

> I agree that this is a better way to combine conflicts than "there are
> conflicts A and B," while preserving combinatoric possibility. My main
> question is whether higher order plot abstractions should pre-exist (e.g. as
> templates or targets or options) in a storytelling system, or if they should
> be regarded as something that arises out of a general process of conflict
> nesting. In other words, should they be input or output?

I do not understand you here, even if I substitute "higher-order events"
for "higher-order plot abstractions." It seems to me that in order to
tell the player anything, we have to have some kind of event templates.
It is certainly possible (not sure how desirable) to tie them tightly to
certain subplots, i.e. "event X can only occur as a part of subplot Y,"
but I cannot see how subplots could produce the events without templates
for them being specifically authored. I guess I'm just missing your
point here.

> IF "X endeavors to accomplish B" is A Good Plot
> AND "Y endeavors to prevent A" is A Good Plot
> AND a whole passel of filtering conditions regarding X, Y, A, B, and the
> relationships between them are true
> THEN "X threatens to do A if Y does not do B" is A Good Plot.
> 
> I think this rule IS valid (but not usable), as long as we leave the passel
> unspecified. Specifying the conditions so as to make the rule useful without
> too far compromising its validity is the challenge.

True. Of course, the problem is eased by the fact that we do not have to
make it valid for all conceivable actions A and B: we only need to
consider those that we, the authors, create for this dream. I hope that
in the context of a specific dream, the author can make a good guess
about what the rules should be, and playtesting can help fine-tuning them.

Earlier, you say:
> Unfortunately, [the rule] obviously not valid as stated so far. If X is
> an astronaut and B is landing on the moon, a plot in which X threatens to
> shoot Y if Y doesn't land on the moon isn't going to be comprehensible, let
> alone make the bestseller list.

Although I do agree that the passel of filtering conditions will very
likely be needed, I would have defined the first two conditions
differently in the first place (for practical considerations):

IF "X wants B to happen, which Y is able to do"
AND "Y does not want A to happen, which X is able to do"
THEN "X threatens to do A if Y does not do B" is A Good Plot.

This seems to be more useful anyway, because we only need to consider it
for any A and B that can actually be done. It would also have avoided
the case you describe, because "X threatens to shoot Y if Y doesn't land
on the moon" would only be considered A Good Plot if Y were able to land
on the moon, and X wants *anybody* to land on the moon, not themselves
specifically. And then again, the plot may make sense-- X and Y are on
some space vehicle, only one of them can land on the moon, it's
dangerous to try to and they're both afraid of it, but X feels one of
them has to land on the moon because they put the reputation of their
country over anything else.

> I believe this challenge is similar to or perhaps a bit greater in magnitude
> than the familiar challenge of determining whether an event sequence "A does
> X, and in reaction B does Y" is valid.

Hm. Of course, what I stated above not *enough*-- obviously, at least we
need to put the relationship between X and Y, and the desirability of A
and B into the equation. But I think that like we can solve the problem
of event sequence by a realtively simple weighting equation (in
Erasmatron), we can similarly do so in this case; we do not have to take
every possible influence into account, a small number of major ones
should suffice.

> (Note that we need not only a possibly very large set of filtering conditions
> for this rule, we also need a comparable rules for every type of higher order
> plot element / plot nesting relationship).

Hm-- yes. This may be a problem. On the other hand, if we use nested
subplots, the number of action templates that can actually be taken in
the context of any specific subplot may be limited (even though the
number of ways to instantiate these action templates might be very large).

> Taking this approach offers the possibility of tackling the problem of plot
> construction by sheer brute force. [...] In the history of AI, being able to be
> solved by sheer brute force has always been the key requirement for a problem
> to actually be solved.

True. -- Do note that our problem is not generating a Great Plot; the
requirements for an interactive dream's plot seem to be much more lose
than the requirements for a linear story's plot. But then, of course,
the AI research in plot generation is far away from even considering how
to create a truly well-crafted plot.

- Benja