Michael Mateas is a prof and video game designer who's managed to make it in to my pretty nebulous list of favourite AI Researchers. I doubt you recognise his name unless you ever bother trolling game studies blogs, but it's likely you've come across at least one of the games he's helped design. Facade in particular garnered a lot of academic acclaim for being the first successful AI-based electronic narrative: becoming a bit of a pleasure sock for the likes of Kotaku and even being called the most important game in the last decade by Ernest Adams. It's actually a brilliant little project, but my interest in it is specifically in Mateas' idea of procedural authorship.

As an academic Mateas suffers from an affliction of making up bogus terms that I don't understand. But if you're lucky enough you're usually able to find some sort of material that explains it all well enough.

Here's a video that's over an hour long that you won't watch:



Mateas takes a position that tries to reconcile player-agency with authorial control in video games. This follows the common ludological conceit that narrative is fundamentally incompatible with the player's ability to have control over his actions or the plot arc, and is instead railroaded into doing what the designer has dictated. In other words, any kind of agency in a video game narrative is always in one way or another going to be predetermined. Think about video game flavour text which basically tries to add depth to the game but in a superficial way that is based on triggered responses.

So Mateas' argument is that explicitly scripting actions creates an undynamic game experience. Instead he opts for introducing situations and NPCs that react dynamically to player action based on the reactive planning language ABL. ABL programs are organised as a collection of many different behaviours which are activated based on how naturally they fit with the actions the user takes. So a behaviour like 'yawn' will be scripted into the game but when, how, and why it is activated is based on the actual dynamic context that is being generated by player interaction.

In his words:

"what the architecture should do for you is allow these behaviours to mix and match richly and for multiple behaviours to be active at the same time, and to get a behaviour that makes sense out of the character that doesn't look like dithering or thrashing....for instance the system is figuring out that [grace and trip's] overt emotion is happiness but [they're] covertly they're really upset so it needs to create a strange smile but with this look of surprise when a certain topic is brought up, and so on"

So authorship in a game becomes entirely procedure based; rather than explicitly dictating the actions of the plot, the real role of the author here would be to create framework while the user propels the actual story. It brings up some interesting questions about the nature of content versus form, ludology and narratology, and even the role of the author.