#tint2016

Time for yet-another conference write-up.

This week, I’m at ICIDS (The International Conference on Interactive Digital Storytelling), and like any good conference (well, at least, “like most conferences I’ve attended”) it opened up with a dazzling array of workshops.

Of those, I spent my entire day at TINT - The 1st Workshop on Tutorials in Intelligent Narrative Technologies. And, in a nutshell, it was super cool. The workshop consisted of four different talks on four different technologies, all of which were given to us to play around with, as well as walk thoughts of both the design considerations and the primary use cases from the mouth of the designers themselves.

For more information, you can check out the workshop’s webpage here

First, we had Stephen Ware’s planning system, Sabre. The talk was super fascinating - it started off with a description of what narrative planners look like and how the operate to construct narratives that don’t just short-circuit the interesting paths that stories often take. Since, in general, planners are trying to find the easiest way to string actions together to met a particular set of goals, the easiest thing to do is to create a non-character entity within the system (often represented as the “author”) who has goals related to making the story more interesting, as well as the goals that would conclude the story. The other important feature within a narrative planner is explanability - it’s super important that whenever a character takes an action, we should be able to backtrace through the state to explain what caused them to take said action. Sabre itself is an neat little environment, though I stepped out briefly and when I came back, I had missed most of the explanation of how to operate the demo. Documentation of the commands that you can enter would have been invaluable.

Next up was Chris Marten’s Ceptre system, which strives to be somewhere inbetween the hardcore AI planning side of interactive narrative and the easy-to-author side. It uses a formal logic system of making statements about the world, and any actions one takes make changes to the state by immediately replacing all of the preconditions of an action with the postconditions. So, if my character is in Room A, there can be a rule that replaces that with the character being in Room B, symbolizing that the character moved from one room to the other. This tutorial did also open up with my favorite example domain of all time - the OTP generator, which just given a list of characters, pairs them up.

After lunch, we had back-to-back talks from UCSC peeps. First was Kate Compton’s Tracery, which honestly? I don’t think I’m going to be able to do Tracery justice here. If you haven’t heard of it, it’s a light-weight javascript library for context-free grammars, and there’s a Twitter Bot service (Cheap Bots Done Quick) powered by it. It’s really expressive, really easy to use and if you want to know more, I highly recommend poking around at http://tracery.io/ for more information.

Finally, we had James Ryan’s Talk of the Town engine, a system that currently simulates a small town in middle America from the mid-1800′s all the way up to 1979. It’s a fantastic piece of technology, firstly, and it’s the backbone behind Bad News, as well as several of James’ other projects. But where it really shines is what you get when you can poke at a simulation that exists on your machine - James has baked in SO many functions for getting at the exact specific kind of thing you’d want to see out of this sort of sim. Once we all got it up and running on our machines, we spent most of the rest of the session exploring the life and times of one particular character in the town on the main computer, Yvette. It was a super fun session, and I really look forward to poking around at this system in my (obviously infinite as a grad student) free time.