ICIDS, Some Thoughts

ICIDS is an academic conference in interactive storytelling. This year it was held in Singapore, and I was invited as a speaker, which was awesome. I spoke about Versu and Blood & Laurels. Graham Nelson accompanied me, and though he was sightseeing for some portion of the trip, he did come along to one of the workshops, which tackled story modeling and authorship; and we co-guest-taught Alex Mitchell’s class in interactive storytelling at the National University of Singapore.

I found the whole experience pretty fascinating. ICIDS is devoted to a lot of the same things that we discuss in the IF community: how to make interactive stories, how to build tools for authors, how to make use of procedural techniques to do things that stories haven’t mostly tried before, how to gauge player/reader responses to interactive story experiences and learn to make better ones.

At the same time, it’s a very different world. Most research funding goes into designing systems, which means that there are a plethora of tools and techniques here, many of them very much more exotic than the tools we typically use in the IF community. On the other hand, not much funding seems to be available for building things with those systems — which means that a lot of the papers abstractly discuss the virtues of particular procedural approaches without ever having applied them to a single full-scale interactive story.

Likewise, standards of evaluation are different. Having read through the proceedings before my talk, I was keenly aware that I hadn’t done the kind of study of player response to Blood & Laurels that academics were used to seeing — namely, getting a sample group to fill out a questionnaire about their feelings about the game, then providing a scrupulous statistical analysis thereof — and instead was relying on the rather less uniform data one gets from QA feedback, pre-release testing, and reviews.

Moreover, as is often the case with academic publications, the proceedings are published in a way that makes them very difficult to get at if you’re not yourself hooked up with a university: they’re put out by Springer, and buying access to a single individual article will run you £23. (From a UK address, I don’t see what they’re charging in the US, but if it’s the same amount at current exchange rates, that would be $36 right now.) That’s a lot of money to spend on spec to look at an article that might or might not prove useful to your hobbyist pursuit, might or might not be written in terms you’ll understand, and might or might not require additional background reading in the form of further Springer-locked articles (ka-ching!).

There are some points of contact and some individual attempts to reach out (among them the fact that I was invited to this conference at all). Long ago, I wrote about some work by Nicolas Szilas, and he commented to discuss it — inviting me, in fact, to help with his project, though I didn’t have the time to be of any use (a familiar refrain). Likewise, some people definitely cross over between the two worlds: Clara Fernandez-Vara, Nick Montfort, and many graduate students and professors at UCSC, notably Aaron Reed. Nonetheless, I think there’s room for more, and more productive, connection between that world and the IF communities. Quite frequently we have speculative conversations in IF about what could be done with more/better/different tools, but we aren’t aware of the relevant research; conversely, in my biased way I feel that experienced interactive fiction authors might have some very valuable insights into the tools being created.

In which spirit, here were some highlights from my trip that might interest folks here:

Interactive non-fiction. As our recent theoryclub session revealed, the IF community doesn’t have a great deal of work in this area (or else the people who know a lot about it happened not to be present). Quite a lot is going on here, though. Notably, William Uricchio (MIT) gave a talk on interactive documentaries that really deserves its own blog post; there were also talks on CHESS, a system that acts as a personalized guide to a museum space by learning the user’s interests and bringing her attention to suitable supporting media, and SPIRIT, an augmented-reality guide app that ushers users through historical sites.

Procedural approaches to text and reporting. James Ryan (UCSC) presented a paper on Combinatorial Dialogue Authoring that I found particularly relevant to my own interests: it deals with the problem of creating dialogue suitable for the many different situations that arise in Prom Week. Prom Week uses social simulation to track the relationships between a number of high school students, and allows the player to make characters interact in various ways; it remembers not only how people feel about one another in the present, but also the content of past interactions. As a result, the Prom Week world can get into many many many possible simulation states; even with a large body of hand-authored dialogue, there isn’t always a good dialogue fit to represent exactly the situation that the world has now reached. That means that what the player experiences is a representation of the world model that is less exact than the model itself; much of the strength of the simulation is hidden from the player. The combinatorial authoring approach attempts to resolve this by tagging a bunch of human authored dialogue to indicate what information it conveys, then have the computer create a number of new dialogue exchanges by recombining dialogue elements. (This requires tagging not only about what the dialogue means but about the types of speech acts involved: is this sentence a question and does it have to be followed by an answer? etc.) The newly generated dialogue can then cover a wider range of situations than the hand-authored dialogue alone, making it more likely that the reader will receive a very specific and customized depiction of the situation in hand.

Along the same lines, there was a talk on changing narrative style in Scheherazade: Scheherazade is an ongoing project at Georgia Tech which collects a large number of human-generated stories about the same topic (e.g., “a bank robbery”) and then builds general systems for telling stories of that kind by recombining and rephrasing sentences. I find this approach somewhat problematic in that it seems too detached from any model world, and too difficult to guarantee (as a result) that inconsistencies won’t arise in the narrative; for that reason I’m a little more compelled by approaches that try to do a good job of narrating to the player a series of model-generated events. But I could be wrong about the productivity of this approach, and in any case this particular paper was about trying to extract word connotations from WordNet so that the same story could be retold with more positive or negative vocabulary, with more or less emotion, etc.; which strikes me as fairly interesting, if very distant from most of the things that modern IF is mostly doing. For bonus points, they do make their research papers available for free, as linked above.

Human-computer interaction studies. Kamil Kamysz and Marcin Wichrowski showed a children’s book app in which the top and bottom halves of the page can be altered interchangeably. This works a bit like those old board picturebooks where you can, say, match any top half of the head with any bottom half, but (being an app) it goes a step further: the bottom text changes in response to the chosen top half, and objects can be dragged from one area to affect objects in the other area. This piece lives in an interesting space between the affordances of children’s picturebook and the affordances of graphical adventure, and it was also just a beautiful piece of design. (Sadly I can’t find any pictures to put here.)

There are several other things that I’d like to write up from the conference that really deserve their own posts; more to follow as I have time.