May 7, 2004

Teaching Interactive Narrative

by Michael Mateas · , 12:01 am

This Spring I taught Interactive Narrative. In this class, through a mixture of readings and projects, we survey the landscape of interactive narrative, examining the theoretical issues, debates and design issues that arise around different conceptions of interactive narrative. As I’ve discussed previously, the class is organized around technical genres (e.g. interactive fiction, author-based story generation, interactive drama), where a technical genre consists of a community of practice (history of work and criticism) organized around specific computational and design commitments.

The class is heavily project-based – 6 weeks of the class are spent in two 3 week design cycles, including in-class critique, in which students design and implement an interactive narrative. For these projects, students are free to explore/invent any form that interests them, as long as they can articulate in what sense it is interactive and (harder) in what sense it is narrative. We actually look at works along four design dimensions, in terms of interactivity, narrativity, segmentation and representation (as I discussed earlier).

I do spend some time in the class exploring the ludology/narratology debate.

I find this debate quite useful as a way to force people to think about what narrative means in the context of interactive work. It is far too easy for “narrative” to become so broad, so general, that it can mean everything and anything. In asking “In what sense is this work narrative?”, I don’t expect there to be a single answer that covers all cases, but I do want people to be able to articulate an answer in specific cases. And sometimes the answer is that a work has nothing to do with narrative, or invokes narrative so weakly that there’s little to be gained by engaging the work (either from a design or analysis perspective) as narrative. In order for “interactive narrative” to mean something, there have to be things that are not interactive narrative. Despite dissatisfactions regarding the ludology/narratology debate felt by both sides, I have found the debate helpful in forcing me to clarify, both in my own work as a theorist/practitioner and as a teacher, what the devil “narrative” could possibly mean in the context of interactive work. The debate is a corrective to a pie-in-the-sky, feel-good, “the world is full of stories; everything is narrative” attitude that is vague, sloppy, and worst, from the viewpoint of a theorist/practitioner, useless.

Relative to an earlier discussion about terminology for interactive “story-like” work, I use “interactive narrative” as an umbrella term to cover the entire space, and historically determined terms used by communities of practice for specific technical genres. So, for example, even though “interactive fiction” could potentially be applied more broadly, I use it only to refer to text-based work that one interacts with using a parser (Zork-like in form), since this is the term employed by the community of practice organized around this technical genre. There’s no reason to force some specific label to cover the entire space. Incidentally, one of the technical genres I don’t currently cover in the class is in fact “interactive storytelling”, the term that caused Andrew to start the earlier discussion. Interactive storytelling systems support the user in creating their own stories to share with others. The most common example are the digital photo systems that supposedly allow users to tell stories with their photos. Unfortunately, such systems tend to be little more than electronic scrapbooks – the sense of narrative is weak enough that I currently don’t include a discussion of these systems in the class.

One of the challenges in teaching the class is the breadth of material covered. Since we only spend a week or so on each genre, the students gain depth through working on the two projects. But gaining enough depth to critically engage a specific form is difficult. For example, both years I’ve taught this class, some students have chosen to do Inform-based interactive fiction for one of their projects, learning Inform on their own, with out-of-class help from me if they need it. Besides picking up Inform fast enough to do a three week project (something the students have done admirably), creating an informed (no pun intended) work of IF requires having played IF works fairly broadly. Students tend to have ideas that come up over-and-over again in IF, such as the unreliable player character, ironic commentary on the kleptomaniacy of the traditional text adventure, various forms of metalepsis, etc. To do these ideas justice, ideally they would have played the umpteen works that treat these various themes before writing their own work, but there just isn’t time (I try to point out one or two relevant works to students on a project-by-project basis).

Another challenge is the lack of tools. Many students in the class have low or moderate coding skills. Besides impacting their project choices, it limits what I can do with the smaller exercises. For example, ideally, for the morphological analysis exercise, I would have students implement their analyses as a story grammar. In class, we look at Propp’s grammar as an example. I’ve implemented part of Propp’s grammar in Prolog grammar rule syntax, which, besides revealing the woeful underdetermination of Propp’s analysis, allows me to talk about more modern approaches to story grammars, issues of hierarchical structure, issues of grounding the leaves of the grammar (how do you actually produce representational elements when the grammar grounds out, perhaps by invoking a simulation engine, etc.) and so forth. But even though the Prolog grammar formalism makes writing decomposition rules very easy, a prolog compiler and an emacs buffer is still not an appropriate tool for a one week exercise for students who are not all extremely comfortable with programming. Similarly, both years I’ve taught this class, students have become interested in the Universe model of story generation (author plans), and have implemented interactive works that make use of this model for their second projects. But much of their time is spent implementing an engine that is a weak copy of the Universe model. Ideally, I’d have a nice graphical authoring environment for both story grammars and author plans, allowing students to focus on the more interesting issue of how such models can be employed within interactive experiences. I have some tool-building plans for next year that should begin to address this issue.

4 Responses to “Teaching Interactive Narrative”


  1. Jill Says:

    Oh, I recognise the woes of not having enough time to do it PROPERLY. Learning’s slow though, as I realise watching my daughter learn to play the violin or my students learning CSS and about what kinds of things might be done on the web – sure, they know so much more now than in January but there is SO MUCH more to learn. I wish I were teaching a continuation class this autummn. Well, not really, I’ll have fun teaching something else this autumn, but it does feel like unfinished work helping this spring’s students here and no further.

    On another matter, though: I’d love to see the grammar engine with Propp.

  2. Christy Says:

    Jill said:

    ‘On another matter, though: I’d love to see the grammar engine with Propp.’

    Me too!

  3. nick Says:

    I read your post with particular interest, Michael, since I’m an alumnus of the Interactive Narrative course, back when it was at MIT and being taught by Janet Murray, who of course brought the course to Georgia Tech. The course seems to have survived Janet at MIT, although it looks like it is currently on hiatus. Clearly the class is still going strong at Georgia Tech, though.

    The class was pretty heavily project-based when I took it, although the projects dealt with interactive narratives and precursors that weren’t on the computer. We had a Propp exercise, too, although not one that was as formal or implemented in Prolog.

    much of their time is spent implementing an engine that is a weak copy of the Universe model

    There is an alternative to tool-building that could help, although tools like Character Maker can be a boon in class. Why not let students hack on and expand some free or public-domain system, instead of laboring to weakly copy it? Of course, the trick here is to have such a system on hand.

    It’s interesting that you define certain types of interactive narrative based on communities of practice rather than formal means. Interestingly, there seem to be a lot of people who do poetry generation, but, as far as I can tell, no community of practice there, even though some nice work has been done. Why don’t Jim Carpenter and Eric Elshtain get to hang out at conferences or on a newsgroup or MUD? (Or blog…) Sure, Gnoetry is the freest of free software and Jim codes in C#, but they still would have lots to discuss…

    Back on the issue of the Interactive Narrative course, though, I’m glad your dealing with it capably and hope that tool development work goes well for the next course. And, hooray for the students who use the course as an excuse to write interactive fiction in Inform. I think I was the only one to do so the year I took the course at MIT.

  4. Michael Says:

    On another matter, though: I’d love to see the grammar engine with Propp.

    Currently my little Propp grammar is intended as a lecture aid, something for me to talk about and demonstrate, but not necessarily very understandable on its own. But each year I do improve it a bit. The next time I pick it up, I’ll take it to the point of being publicly releasable. The point of the grammar is just to introduce the idea of a story grammar, and demonstrate the limits of Propp’s analysis from the point of view of building a computational story system.

    I read your post with particular interest, Michael, since I’m an alumnus of the Interactive Narrative course, back when it was at MIT and being taught by Janet Murray, who of course brought the course to Georgia Tech.

    Yes, my course derives from Janet’s course, though I developed a new syllabus. Janet’s course is organized around narrative forms (the ones she identifies in Hamlet on the Holodeck – e.g. multiform narratives, rhizomatic narratives, etc.) rather than technical genres (e.g. interactive fiction, author simulation, etc.). Narrative forms appear in my class under the narrativity design dimension, as we discuss the design commitments made by various technical genres. As you mention Nick, Janet tends to focus on non-computer-based interactive narrative and precursors, while I focus almost entirely on computer-based forms. I tend to stress how different technical commitments (e.g. IF frameworks like TADS/Inform, template-based stateless characters like Eliza vs. stateful, reactive planning characters, various approaches to AI-based story generation, node-and-link hypertext, etc.) strongly influence narrative design choices. Through discussing works in class, plus design critiques, we pull out the common issues in the design of interactive narrative that tend to cut across technical genres.

    Janet and I have been talking about splitting the class into two classes, one that is close to the way she teaches Interactive Narrative, with a focus on non-computer-based work and general design issues, and Intelligent Story Systems, which focuses solely on computationally intense approaches to interactive narrative, and would require programming ability as a prerequisite.

    Why not let students hack on and expand some free or public-domain system, instead of laboring to weakly copy it? Of course, the trick here is to have such a system on hand.

    That is indeed the trick. To my knowledge, there are few (or no) public domain systems for experimenting with interactive narrative except for the tools and languages available in the IF community. If people know of public domain code frameworks available, please post a comment. In class I do discuss the Talespin story generation framework, and demonstrate Warren Sack’s reimplementation of Talespin. One of my students last year did a project in which he modified Talespin to have it generate stories in the anime teen romance genre. But teaching himself lisp (with some help from me), plus Schank’s conceptual dependency framework, was quite difficult, and not something I could assign as a required assignment, especially with non-programmers in the class (but this could be a very nice assignment for the Intelligent Story Systems course…). Incidentally, in his project he rediscovered what we discussed about Talespin in class – that simulated characters pursuing individual goals does not, most of the time, create stories.

Leave a Reply

Powered by WordPress