I usually manage to avoid writing about AI – mainly because it makes me think about work. Work is fun and all yet, inevitably, still work (yeah, I know, I’m really fucking profound in the morning).
But I’ve been meaning to mention this one poster from AIIDE ’07 (sadly, didn’t get to go myself): Player Autonomy vs. Designer Intent: A Case Study of Interactive Tour Guides (via AIGameDev).
The paper’s authors use a type of Markov decision process to model & predict how users move through a museum. That’s not an exact analog of a game environment, but it’s got some useful applications, and allowed the authors to specify a statistical model of how users might navigate the space (some have specific goals of seeing certain exhibits, some don’t, some have seen some exhibits already, etc.). The goal is to improve how many users are able to see the exhibits they want, meet a tour guide defined sequence, and reduce congestion, which correlates well to drama management.
A lot of papers specifically on drama management tend to have “oh hey we should test how well this actually works” under the Future Work section. The neat thing here is, using a more straightforward example, the authors found a sweet spot with AI mediating between player goals and designer defined goals – which tended to improve the end experience over either by itself.
And while it correlates with drama management in a single player game, it also suggested something to me I hadn’t really thought about much before – the idea of drama management in an MMO (where something like congestion very directly relates, and it’s not just about balancing the experience of one user’s goals, but lots of potentially conflicting users).