The Monomyth vs. the Automyth

I subscribe to the GameSetWatch RSS feed, and generally flick through their headlines on Bloglines (an RSS aggregator that I thoroughly endorse, especially the beta version) in a spare moment or two each day. Most of the content leaves me quite cold, but this article actually caught my attention!

I've read around half of Joseph Campbell's book, The Hero with a Thousand Faces, but got bored and bogged down in his continual obsession with his Speilbergian father-son conflict as the driving force of all narrative. (Or is that perhaps, Speilberg's obsession in his movies with Campbellian father-son narratives?) His 1940's writing style doesn't help, but then Christopher Booker's 2005 re-tread and elaboration of the material, The Seven Basic Plots (or here for the US), suffers from the same obsessions, is nearly twice as long, and could have done with some serious editing down to reduce redundant repetition. (Mind you, I did get 80% of the way through this one)

So, I've not read every last word on the subject of the monomyth, or the seven basic plots, but I believe I have an amateur understanding of them. (Yes, go on, my opinions are incomplete and therefore entirely invalid... feel free to go read someone else's blatherings instead).

(note to the reader that chooses to continue, where I use a male pronoun, please feel free to substitute a female one if it helps you)

Quick summary: hero in the 'normal' world is challenged to go beyond the boundaries, overcome 'evil' (the father) in some way, returning changed to the 'normal'. There are, of course, many more details to monomyth, and the seven basic plots (quest, rags-to-riches, overcoming the monster, voyage & return, comedy (ending in love or life), tragedy (ending in some form of death), and rebirth) have more subtlety than that, and the article at GameSetWatch gives a more complete summary of the basics.

However, they mostly miss the point that elements of good story telling also involve subverting the monomyth, or the expectations of the basic form of the plot chosen (consciously or otherwise). For example, the hero might return from the experiences unchanged in himself, but the 'normal' world might instead have been changed as a result of his actions.

Anyhow, the point of that article is to ask why, if the monomyth is so well explored by Campbell and his acolytes, has no-one actually built an 'automyth' - a monomyth generator. This is the intriguing bit. It's all very well stating that there's an algorithm for a story, and even that one can subvert (invert?) or skip one or more, or most, of the 17 steps in the algorithm, but without a detailed language to express a world model, or two (possibly more) in which to run that algorithm, automation is just a dream.

How does one tackle filling in the steps in the automyth algorithm? Perhaps by giving the machine a library of options:

• people: farm boy, new advertising executive, grizzled mountain man, washed out detective
• places: Alabama USA, the Shire (or an analogue of), a moon-base, a small shop in London
• challenges: alien invasion, disease, death of a parent, financial ruin (personal or global)
• monsters: the boss, an out-of-control robot, the Devil, the judge(s)
• assistants: the girlfriend, the grand-parent, the old mage, the bestest buddy, the advocate
• ... and so on

But does this actually achieve anything? Can we build a grammar with which to join these elements? Do we need to?

I think what we're looking at here is a new expression of the Turing Test (a test where one person cannot distinguish which one of two subjects is human, and which one a computer, by means of asking questions). And if we define a world, or set of worlds, within which the automyth machine will operate, we have a limited Turing Test (the questions are limited to a specific subject).

At which point, can we talk about applying this to ludonarrative generation? Certainly we could build a quest generator (a la Diablo II), but that's only a subset of the challenge. Can we really express enough full plots. I think not, at least, not until we have a way to describe the entire world to a computer, give it a way of imagining alternatives, then (and this is the really difficult bit) giving that computer a way to judge which of those alternatives could actually hang together as an alternative. This is such a big job, I think I'm off to do the laundry instead.