I watched Ken Levine’s recent GDC talk with great interest. I’ve spent a lot of time thinking about how designers model (or don’t, or might) social interactions between player characters and non-player characters, so Levine’s idea for what he calls ‘narrative lego’ is right up my alley. Put simply: Levine envisages a system in which NPCs each have several ‘passions’, randomly assigned, and will react to the PC based on how they relate to these. If Mr. Orc doesn’t like elves and the player gets all friendly with elves, the PC will drop in his esteem, and so on. As the PC reaches certain levels of repute with the orc, he exhibits certain behaviours. In Levine’s example, the orc blacksmith might grant a discount.
In a sense, RPGs both digital and analogue have been doing this for years through scripted events. Levine takes the (well, a) logical next step and attempts to formalise these behaviours mechanically. What he is constructing is therefore a social simulator rather than a scripted narrative.
While I find this all fascinating, I have a couple of reservations. Firstly, I think that the lack of direct association between an NPC’s ‘passions’ and the way they respond to the PC is exacerbates a problem a lot of RPGs already have in abundance. To give a concrete example: the infamous ‘kindness coins’ approach, epitomised by Bioware’s games, in which an NPC designated as a love interest will sleep with the PC once the PC rises sufficiently in their esteem by giving gifts and/or making sympathetic noises. This system is simultaneously creepy, unbelievable and (dare I say it) a symptom of the sort of market-esque, quantitative approach to human behaviour with which the genre is already replete. See also: the good vs. evil meter, where the PC acquires ‘goodness’ and ‘evil’ as if they were discrete currencies (have I just blundered into some kind of market/metaphysic/ludic nexus?).
Consider an alternative approach (yes, I’m about to do some armchair game design — but I’m definitely going to use this in my current pen & paper campaign). Let’s say that each passion is tied to a particular kind of response. So, yes, an NPC might still be designated as a possible love interest, but ‘romance’ becomes a particular passion with its own set of stimuli and responses, which might interact in unexpected ways with other passions. Just because someone wants to sleep with you, doesn’t mean that you’re friends, and vice versa. Some NPCs might even be torn between miltiple passions in the form of competing allegiances — love or duty, friendship or loyalty? — and be led to conflicted forms of behaviour, rather than the grey mush that would otherwise result from competing quantitative modifiers (Levine identifies this as a problem in the talk, but his suggestion for a solution is essentially, and appropriately, mathematical).
I also think that Levine’s emphasis on procedural generation goes too far. If the setting itself doesn’t contain interesting conflicts, it may result in yet more grey mush. This is partly a matter of taste and of politics, but a merely random (meaning personal, rather than ultimately social and structural) set of conflicts just doesn’t interest me that much in a story. Which is to say: I don’t think you can do good procedural worldbuilding.
Finally, there is the ‘do ut des‘ approach to human interaction. While a model of behaviour that amounts to making offerings (or performing quests) in order to elicit a desired response might work rather well for relationships with the deities of a polytheistic pantheon (so much so that it I’m amazed that fantasy RPGs never do this — what does it say about the genre that it is full of upside-down worlds where people are like Greek gods and gods more like people?), but are human interactions that simple? The task-oriented approach to behaviour here is probably both overly tied to a particular concept of human agency (of which RPGs and realist literature partake) and insufficiently interested in the complexities of human affect. Which is only another way of stating the obvious fact that if you want to create a system of human behaviour, then you need a model to work from.