Honours project for BCS; Generating organic questlines through icon based dialogue composition.
In this prototype, you talk to villagers. You start with very little information, but collect information and exchange information with the intention of befriending villagers and obtaining quests.
It's very difficult to experiment with right now. Most of the research was focused on learning about how we apply bias to information during discussions, how that bias plays out when communicating with people, and how we can represent information and biases toward information in games.
The problem that triggered this research is what I think of as the "Sith Lord" problem. If you make a dialogue choice in a game, it may not accurately reflect your intentions. You can't really lie unless the game presents you with an option and then specifies it as a lie. To overcome this, I wrote a dialogue system which allows you to exchange information however you'd like.
Unfortunately the prototype can't really demonstrate a lot of this research. It can generate quests, but there's little else to do than to go around having conversations with people. Given that the focus was on the research and less on the prototype, the game feel just isn't there to make conversations feel fun so you're not likely to experiment enough to form relationships with the townsfolk.
Over the years I've kept this project in mind and thought about ways to bring this emotional information exchange out in a more interesting light. I've been working extensively with word feel and emotional sentiment, getting to understand various models of emotional classification, and I think there's some great work to be done in creating freeform dialogue systems and NPCs with memories, personalities, and opinions.
This is just one experiment closer to trying to emphasize charisma and social engineering as a mechanic in single player experiences.