The other night one of my former coworkers, Kent Hudson, gave an IGDA lecture on player-driven stories. It was a fascinating and inspiring talk, but I did have a couple of thoughts that might be contrary to some of his conclusions.
The slides for his talk, with great annotations, are available here. Or if you have access to the GDC vault, he gave mostly the same talk earlier this year. As a quick summary, though, he talked about Self-Determination Theory and related the pillars of Autonomy (for which he substituted Agency), Competence, and Relatedness to player-driven narratives. He discussed the current problem of game narrative, ludonarrative dissonance, and cheap and easy ways to improve game stories. He also outlined a theoretical system to read in a player's important ludic decisions and dynamically adjust the narrative in response, creating a story that is tailored to the player.
Kent argues that without Relatedness (the happenings in the game being tied to a recognizable, relatable story arc), games with Agency and Competence just offer "systems soup", and that doesn't lead to satisfying player-driven narratives. I'm not entirely sure I agree with that assertion.
The two games that generate the stories I most like reading or hearing about are EVE Online and Dwarf Fortress. EVE's stories are amazing, and range from "The Great Scam" to the "Great War" and the "Second Great War" (in space, everything is great) involving this tactical masterstroke of corporate espionage. I don't play EVE, and I don't plan to, but reading these stories is really exciting and fascinating to me. The thing is, EVE is systems soup. There is absolutely no intent by the designers to author any sort of narrative at all, not even an emergent systemic narrative as Kent proposes. And yet it has these amazing stories.
That could just be because EVE is an MMO - everyone involved is human, and so the stories are just human internet drama with added spaceships. But consider Dwarf Fortress, with stories like Boatmurdered and Bravemule. Dwarf Fortress is like the Campbell's Extra Chunky Systems Soup, now with 50% more vegetables. It is basically just a giant fantasy world simulated at an amazing level of detail, and stuff only works because of the emergent interactions of all these delicious systems. There is no narrative arc at all, except the inevitably unsuccessful struggle for survival in a harsh malevolent world. There isn't a player avatar, just a cast of procedurally-generated dwarves with certain skills and likes and dislikes that adapt to their situation. And yet something like Bravemule emerges from that.
I would argue that this happens precisely because of the lack of explicit narrative. People are very very good at finding patterns in things, especially patterns relating to people, and even when no such pattern actually exists. That's why we see shapes in clouds, or a face made up of punctuation: ;-) And I think that's why we can see these epic stories in Dwarf Fortress - we see just enough in the interactions of the systems that our brains superimpose the epic-scale human drama on the lives of the dwarves (which, much like them, are nasty, brutish, and short)*.
One example of a cheap narrative choice Kent brings up in his talk is the famous companion cube from Portal - simply with some dialog and a texture, Valve created an inanimate object that people feel an emotional connection with. But again, almost all the work is being done by people's brains imposing their human-focused interactions and relationships on this virtual inanimate object.
So, minor disagreement one: I don't think "systems soup" games are necessarily** devoid of excellent narrative potential. In fact I think they are perhaps one of the best ways to generate compelling narratives because the player's imagination is doing most of the heavy lifting. It's a prime example of the ELIZA effect.
And along those lines: the arrangement of a few dots and a curve suggesting a face to us is really interesting. What's even more interesting is the fact as the representation gets closer and closer to an actual face, without getting there, the recognition gets better and better - until it plunges into what's called "the Uncanny Valley". I think the same thing might occur with systemic games and narrative - if we push too close to actual relationship modelling and dynamic story-arc creation, we might get something that turns people off instead. It is pretending to be a story, but isn't, and we can smell the weirdness behind it. It will end up like ELIZA - initially compelling but ultimately weird and off-putting. I seem to remember that I've played a game that made me feel that way, but I can't for the life of me remember what it was. Something where the authors had pushed procedural generation too far and ended up with nonsense rather than a believable world.
I think the system Kent discusses in his talk is probably not getting anywhere close to this, but I do think it's important to recognize the power of people's imagination in filling in the story details. All you need to do is give them a framework, and they will hang their own artistry on it. Bulking up the framework might make that easier, but you have to be careful not to stomp the player's own imagination spaces at the same time. Kent obviously has great respect for the player's imagination and its power to suggest stories, so I don't think he's in any danger of running into an uncanny story valley.
*(sadly, I'm not that clever - I stole that joke from Terry Pratchett)
** (Heavily systemic games don't necessarily generate narrative on the scale of Dwarf Fortress either, though)