Home About EIS →

Emily Short on Playable Narrative Systems

Interactive story author Emily Short spoke at UC Santa Cruz on Wednesday, as part of the ongoing Inventing the Future of Games speaker series. Emily, best known for her work on groundbreaking interactive fictions such as Galatea and Savoir-Faire, spoke about her recent work with Richard Evans (AI Lead on The Sims 3) developing an interactive story system centered around character and conversation in Jane Austen’s universe. (Emily and Richard’s company, Little Text People, has recently been acquired by Linden Lab; Emily stressed that the ideas presented in the talk represent work done before the acquisition.)

The problem Emily tackles is one familiar to storytellers in interactive media: how is it possible to have meaningful interactions with other characters without creating an impossible authorial burden of endless branching conversation trees? In other words, how can we build a system that replicates something of the experience of interacting with other people, rather than authoring every possible interaction by hand in advance? The approach taken here is based on “social practice modeling,” tying NPC behavior to an expressive AI engine that understands interactions in a specific social milieu.

Designed to be something you could curl up by the fire and read on an ebook or laptop, the system presents an ongoing textual story that advances in real time as you read it. The player picks a character in the story whose role to adopt—perhaps the detective in a mystery for a very hands-on story, or a minor character (even the dog!) for a more passive experience—and can then take actions in the world based on the current social context and character’s place within it. For instance, as a guest at a dinner scene it would be appropriate to make small talk with those present, eat food, and share flirtatious glances, but not to abruptly rise and leave the house without provocation. Playing as a servant in the same scene, however, one might have very different affordances. (The player has some ability to go outside expected behavior through a pool of resource points.) The player’s available actions from moment to moment are also based on their relations with other characters and their personality: a shy person might have different conversational affordances than a loquacious one. These traits also drive the behaviors of the non-player characters, ensuring they take appropriate actions. Interestingly, the system is designed to support multi-player: the AI controls all remaining characters not played by a human.

Behind this framework lies a complicated map of characters’ desires and beliefs. Each character has a model of things they believe to be true, including opinions about other characters and their relationships with each other, like “Emma is nice but talks too much” or “Mr. Elton is in love with Harriet.” As characters observe other characters take actions that conflict with this model, these opinions are revised (“Mr. Elton is in love with Emma!”) and characters are able to verbalize or question these opinions in conversation. This allows for dynamic conversations that eschew a pre-scripted conversation tree in favor of multiple agents each changing the social landscape with each “move” they make, both advancing and evolving opinions about each other and the plot from moment to moment and scene to scene.

The system also tries to solve the problem of embedding these moment-to-moment actions into a larger narrative structure. For instance, characters are in part defined by traits, but at dramatic moments in a story, players can push characters to change by abandoning traits or acquiring new ones. For instance, a shy character trying to get up the nerve to propose marriage to a girl he loves might spend some of his in-game points to overcome shyness just to propose, or might spend a large amount of it to overcome his shyness permanently, for presumably a much more dramatic proposal. Players have the option of saving the developed version of their characters to play in other stories.

It’s interesting to note that many of the approaches in this project mirror those taken in Prom Week, released this week here at UCSC; Emily commented on these similarities several times in her talk. It’s not surprising, as both of us were trying to solve similar problems of building playable systems from social interactions. It’s great to see powerful expressive systems driving story and character starting to move beyond theory and into real playable experiences—in academia, the indie game scene, and the world of commercial games, too.

About the author:  Aaron A. Reed is a PhD student at the Expressive Intelligence Studio working at the intersection between computation and literature. He has previously received an MFA degree in Digital Arts & New Media from UC Santa Cruz. Read more from this author

This entry was posted in Academics. Bookmark the permalink. Both comments and trackbacks are currently closed.