As we move toward more cooperative and competitive online play, there are some gamers out there who think that artificial intelligence research isn’t as important as it once was. If you take a step beyond the prevalence of horde modes and NPC behavior in MMORPGs, you can see that AI research is not only alive and well, but dedicated to fundamentally changing your relationship to games.

And yes, AI will kill you in the process.

First we need to realize that AI research isn’t contained to simply the study of NPC behavior. Yes, it’s always good to have NPCs walking around throughout a game world, ducking behind cover, executing seemingly random scripts, making their behavior seem lifelike. But AI is so much more than that.

AI research also encompasses systems and algorithms capable of untangling the mass of data left in the wake of gaming. Data mining is big business right now, and part of the issue with it lies within the sheer volume of information that developers are able to gather from the games we play. In-game behavior logs are one of the many sources of data-collection utilized by developers. Anytime something significant occurs in an MMORPG (killing a monster, finishing a quest, joining a group), a time-stamped record is logged; later this log can be parsed by any number of ways (IEEE Intelligent Systems, Vol. 26, No. 1, pg.86. Online here [subscription]).

Log 231: The users have decapitated 4,523 boar in 2 hours 28 minutes

AI comes in because people need a way of gathering and analyzing possible patterns. Utilizing scripts (extremely basic, non-adaptive AIs), developers can find statistically important areas for farming, competition, or griefing. They can then use that data to patch the client, resolving potential issues (occasionally, developers and publishers make that information accessible, too).

The real exciting and terrifying aspect of this increased use of data analysis is that previously, the information would have to be parsed by humans with limited algorithmic assistance. With the other aspect of AI being brought into the equation (the kind of research into adaptive and learning forms of AI), we are beginning to see the capacity for autonomous data analysis and real-time manipulation. We’ve seen glimpses of this tech in games that adapt difficulty based on performance, but what I’m suggesting here is major incorporation of the data into reshaping the world on a persistent basis.

Implicit in this is the prevelance of Software as a Service (SaaS). Along with our nearly perpetual connection to the Internet, SaaS ensures that there is more constant contact with the developers. Games are always getting patched, and we are increasingly moving to an always-connected model of gaming (IMHO: piracy is not the only reason why developers are moving to always-connected models of DRM), meaning that new data is always coming in, adding to the complexities of the data analysis.

Having automated AI analysis (powered by research similar to the disambiguation research coming out of the University of Houston-Downtown), the systems themselves can monitor and update as needed, changing the gaming world to whatever form is best suited for customer-engagement, hours logged, and community building. With Agile and Lean processes being utilized on the back end of software creation, you will soon be seeing teams of workers (skilled in the various aspects of software development, be it graphics, stability, or user interface) assisted by teams of AI algorithms, able to analyze data, and not just mining for general patterns, but actively figuring out what types of patterns are important to begin with.

Yes, this AI can be nefarious, knowing just where to place NPCs in order to log the most player-kills, but there’s another aspect of this that could be important: realtime manipulation of the gamer him or herself.

The sisters were capable of messing with your mind simply by harvesting…

Taking a step back from the research aspect, here is a list of steps that, all combined, threatens to change the medium’s longstanding structure for better or worse. Think of this as a bunch of slides displayed on the screen (similar to the rundown of events prior to the beginning of Red Dawn); it’ll make it seem more epic.

  • Gamers no longer own their game, they pay for a license to access the gaming world.
  • The developers constantly maintain the game, changing the world as they see fit.
  • Games and game worlds persist outside of a contained virtual world, connecting users and avatars through guilds (with communication online but out of game), social plug-ins (for each like on a post, you’ll deal XX damage), and mobile games that travel with us to the doctor’s office and bathroom (it happens)
  • Research into behavior-inducing effects of pervasive computing looks at what causes behaviors to change in a given environment (virtual environments for the sake of this argument).
  • Automated AI are tasked with increasing gamer involvement for the sake of increasing returns based on money invested.
  • The world, the scope of interactions, and a player’s usage patterns are crafted out of real time manipulation of rules, data, and events.
  • The game that you thought was a game played between real players is really just a simple reflection of an increasingly complex iteration of you vs. the computer.

It doesn’t matter if you are in a group of other players; the AI will know that groups comprised of statistically similar users to yours tend to behave in a certain way in response to a given set of instances. Pulling from that range of variables, it will be able to change the parameters, NPC placements, spawn points, battlefield geography, weapon drop rates, or any other sets of things that will bring about the desired outcome.

The AI Director aspect of Left 4 Dead is a fairly obvious example of randomly generating content on the fly.

Our task as gamers? Resist this futile attempt to make us do things that we don’t want to do. Back in the day, the kill screen was our way of extending a bird finger to the AI, achieving a temporary digital immortality (while some creepy guy walked around the arcade, whispering, “kill screen on Donkey Kong.”). Competition between humans was solely on the vanity of high scores (or how quickly one could get to the kill screen).

I am worried that we are moving to that model again; this time the stakes are much higher. Games as narrative architecture is a worthwhile endeavor, one that seeks the creation of spaces that enable story-tellers to communicate their visions. These new iterations of games have the potential to use narrative as a tool for reinforcing favorable traits in gamers: feeding our vanity, making us mindless clickers, and turning us into good little consumers (Chomsky references welcome in the comments).

For me, the fight against these new iterations of AI is a call to arms. It may or may not want your avatar dead (depending on what constitutes creating a “good” gamer), but eventually, it will kill you, even if it’s only in a metaphorical sense of killing your ability to choose how to interact with the game world itself. But remember: even if 99% of users die when certain conditions are met, there’s still that 1% running around, breaking all the rules, and probably smiling. So grab your interface of choice and be the statistical anomaly that the Architect warned of. Defy probability.

Find that kill screen. Before it’s too late.


2,225 Hits