I was talking a while ago with a colleague (hi, Andrea) about some
stuff that lead onto magic, and whether the presentation of
magic-as-science (with repeatable "spells" and implicit underlying
rules, laws, and understandable components such that if you can
*learn* enough, then all magic is deterministic to you) was anathema
to the spirit of magic itself (a capricious, non-deterministic,
creature of whim).
Recently, I was thinking about words, and some of the things my old
english lit. teacher used to say - about how everyone knows how to
play with words, since it's a basic part of learning to speak, and
his parting shot that no matter what poetry you look at, you should
never get too wrapped up in anything but the words. The words, after
all, are what poetry is *really* about.
Like walking, to talk and the ability to choose and to interpret
words is such a basic essential skill to life that it is extremely
rare to find someone who does not have considerable aptitude with
them.
So, I wondered whether you could make a magic system out of this. My
first thought was simply that players had to extemporize verse in
order to cast a spell. Immediate issues:
- Programmatic Analysis: Meter, scansion, and rhyme are all easy
to measure algorithmically. This should be the basis of rating the
spell for effectiveness.
- Meaning: I haven't met a computer yet that could comprehend
freeform speech, but ... poetry has thousands of years of practice
in running roughshod over literal meaning and applying poetic
license. Very approximate meaning (basic vocab and grammatical
analysis) can in fact be determined - even simple keyword analysis
does pretty well. Where this falls down frequently and
disastrously is an excellent example of "capricious" magic ;)
- Non-determinism: what stops you from using the exact same spells
all the time? This is the point that IMHO previous purely
scientific magic systems (e.g. AC1) have most seriously suffered
from - especially that once spells are "shared" OOG, it's pretty
hard to put the cat back in the bag, even with randomized
per-player recipe components etc.
Firstly, the act of memorizing a poem to use as a spell seems
wonderfully in-character for many games - this is something we'd
like to encourage. So, a certain amount of determinism in spell
success would be a fine thing. However, the amount that a particular
spell is used globally within the game should greatly weaken its
effectiveness. Maybe everyone knows a simple spell for conjuring
light: "From the darkness, give me sight / Show me now, bright white
light", but the fact that everyone knows it and uses it often
degrades its effectiveness exponentially, so that it only has
minimal effect.
Secondly, every spell is easily customizable by the speaker,
encouraging people to at least try some variation on the well-known
version. You are not locked-in to a fully deterministic and
scientific system where customization is either impossible or
extremely limited, but instead in a completely open (artistic)
system where any variation should work, modulo the effects of the
rhyme/meter/etc rules.
Thirdly, as noted above, the algorithms for registering meaning,
meter, etc are likely to be deeply flawed - experimentation will
certainly not have the precise effects expected (at least until the
algorihtms are fully understood; fuzzy matching within them (in fact
any randomization) should make that a never-completed process)
Finally, assuming the bulk of the system is working, then you can
play the effectiveness-rating of these spells much like bad-word
filtering in usernames: all new variants get piped past human
eyeballs (customer support), who cherry-pick things to mark up or
down. This is known to work very well where the bulk of the
processing is automated, and the human element is able to do as much
or as little as it wants, and is able to simply increase the
responsiveness / effectiveness of the automated processing.
This is also a good way of dealing with the next issue:
- Art: good poems should be ranked higher, appalling poems should
be ranked lower. Jokes, external references, emotion - all these
should factor in.
Another way of dealing with this is to allow poems to be submitted
(to a deity, perhaps) for consideration. This means they go into a
queue to be read by staff, who will mark them up or down - but it
could take weeks for a particular poem to make it to the top of the
queue and be considered.
- Power: My first thought was a simple cumulative marking scheme,
which built up a single integer ranking for the power of the
spell, based upon:
* number of consecutive lines of verse in the poem
* amount of repetition (both positive and negative effects)
* bonuses for particular patterns (e.g. repeated meter patterns)
* number of synonyms for key-words found (e.g. for a light spell,
the number of synonyms for "light")
This also has the neat side-effect of making more powerful spells
require a lot more time to cast (/say * N), or much more crafting
time beforehand (optimizing the number of patterns etc that are
worked-in to the poem).
Has anyone done anything like this? It's the kind of thing I can
imagine someone has done before, but I can't think of anywhere where
I remember seeing it.
Adam M