Thoughts on Narrative

I’ve recently been thinking a bit about AI in games, and in particular adaptive AI. In our first lecture for ITB016 this week we briefly went over the topic of narrative, and it set me off once again on the ‘adaptive’ train of thought.

Narrative is traditionally fairly straight-forward and linear in style. Even in examples where the narrative jumps around from character to character, or setting, or time / space (talking more generally about narrative here, not specific to games), it at least follows some predetermined path and is essentially set in concrete (in the case of books, admittedly this is pretty standard practice, due to obvious technological limitations). However, with games, the narrative doesn’t have to be set in concrete. Indeed, there are several examples where games offer rich narrative with multiple paths that can be chosen by the user, and several more that at least offer a couple of very minor narrative variations (you kill the kitten, you get ‘evil ‘ ending, you save the kitten, you get ‘good’ ending).

Although games currently exploit their interactive nature to a degree to provide the semblance of interactive narrative to a player, it really is nothing more than a semblance; in many situations you can even play the game completely differently, but take the same paths at key decisions, and end up at the same place.

So winding back around to my ‘adaptive’ train of thought: imagine a game where the narrative itself is truly interactive, where a player’s interaction with the game is one of the main driving features behind the game’s narrative itself. Would it be possible? Would it be playable? Most importantly, would it be fun? Well in terms of possibility, it would obviously require a large and complex AI model to allow for game narrative to firstly adapt to a player’s interactions, and secondly to replicate that adaptation so it reflects in all non-player characters. Playable? Probably largely dependent on several thousand factors, not least of which being the AI system’s abilities, and whether it allows the game to progress in a way acceptable to players. Would it be fun? Well as with any game, this would be subjective. But subjectiveness aside, there are some areas that could definitely make it not fun. For example, if the AI adapted itself into a corner, or the narrative adapted in such a way as to be boring, ridiculous, hard or nonsensical, then it would likely not be fun.

If this was ever going to work, it would need to be a very well thought out model. My preliminary ideas are thus:

• The system would involve several key narrative events (we still want creative directors – leaving that up to AI would no doubt be disastrous)
• All key NPCs and events would be linked in a complex network including social links, and various cause and effect triggers
• Player actions would ultimately be the starting cause, which might trigger ‘vibrations’ along a certain strand of our web of connections, which in turn causes various events and triggers to be set off

© Jeff Burn 2008

This model is subject to several issues, especially in terms of feasibility. However, I think it could definitely be an interesting area of exploration.