Interactive Digital Narrative: Theory

This is part two of an overview of Interactive Digital Narrative: History, Theory, and Practice. See my earlier post for coverage of the book’s history section (and one practice chapter that I took out of order because it felt like it fit better that way).

This time we’re looking at the theory section, which addresses academic approaches to interactive narrative (including the question of what interactive narrative even is).

Again, the section begins with a brief overview from the volume editors, and this provides a fair sketch of the academic debates of the last couple of decades, together with a bibliography of a number of foundational pieces in this space. I might also have listed Jesper Juul’s half-real here, as it provides a readable and persuasive cap to the narratology vs ludology debate.

Although I have enough of an academic background to be comfortable reading these articles, I’ve often found theory discussions of interactive storytelling largely irrelevant to what I am doing. It’s not that it’s misguided, precisely; it’s more that it tends to be spending a lot of time on questions whose answers would not affect either my creative practice or my criticism, or that I don’t find helpful outside the context of a specific work and artistic brief.

I felt this way about some of Gabriele Ferri’s article “Narrative Structures in IDN Authoring and Analysis” — for which these two quotes exemplify the issue:

With regard to narratives, an interactive matrix does not contain only one plot, but it can generate a large number of actual narrative developments that cannot coexist in a single textual instance. In other words, a single textual output will contain a single plot determined by the user’s actions, but the following session will generate a different one if other choices are made. (84, emphasis mine)

…

Having introduced structuralist and poststructuralist definitions of textuality and narrativity, Greimasian semiotics has been useful by allowing a better comparison and differentiation between IDN and other objects. (84)

I don’t quote in order to ridicule. The first of these points comes after some paragraphs of defining terms and unpacking theories, so that it’s a little like reading a mathematical text that goes back to basic principles to explain why and what it means that 1 + 1 is able to equal 2. But if, as a reader, you are primarily interested in being able to count the oranges in your kitchen, then reaching this conclusion that you knew at the age of four may not feel like a great triumph.

The second quote, meanwhile, concerns a very much live question about the distinctions between interactive fiction and other traditional and new media. But when I am having that argument, it is usually not framed as a polite and abstract discourse based in formal definitions and semiotic theory. It is more often a cage match over access to resources, publishing venues, and cultural capital. We fight it on two fronts — vs. traditional media such as books and movies, and vs. mainstream games. There is so much at stake and so much anger around these questions that I now categorically refuse to answer interview questions about, for instance, whether I think interactive fiction “qualifies” as a game.

So: fair points, and abstractly an interesting read, but not one that gives me tools for what I do.

Or here is a passage from Hartmut Koenitz’s chapter “Towards a Specific Theory of Interactive Digital Narrative”:

An approach based on the structuralist and poststructuralist concepts of Roland Barthes (1973, 1974, 1977, 1979), Jacques Derrida (1982), Michel Foucault (1972, 1977), Umberto Eco (1984, 1989), Jean Baudrillard (1983, 1987, 1993), and Jacques Lacan (1977) and their idea of a narrative that is free of the direct control of the author and open to interpretation led to Hypertext fiction works like Michael Joyce’s Afternoon, A Story (1991) and Shelley Jackson’s Patchwork Girl (1995).

Many theorists and practitioners of Hypertext Fiction (HF) understand [interactive digital narrative] as an opportunity to reflect aspects of poststructuralist positions. In this vein, George Landow describes hyperfiction as challenging ‘narrative and all literary forms based on linearity’… (93)

I have never ever thought to myself, “This game idea gives me an awesome opportunity to reflect poststructuralist positions!” I suspect that that is true for a number of IF authors in the rec.arts-originating parser IF community, even those of us who came from an academic background where we read some of the theorists in this list.

It’s not that that community doesn’t talk about questions of authorial control, linearity vs branching, how we find the end of the reader’s experience in a non-linear work, etc. Rather, historically, we’ve tended to talk about and focus on what those techniques accomplished, rather than what they could tell us abstractly about the nature of digital literature.

IF literature is full of discussions about the effects to be achieved from particular structures and procedural strategies, about how agency is presented to the player and to what end, the distinctions of experience between different user interfaces, and so on, as they pertain to conveying interior states of the protagonist, sketching characters, communicating a critique of a social system, or assorted other tasks.

(These links are intentionally a mix of old and recent pieces, reviews and craft articles and author statements, because the point is that this is a long-running mode of discussion in this community and that it is conducted at multiple levels of formality.)

It’s the discussion not just of practitioners, but of practitioners of art that is often about something other than the art form itself. Perhaps that’s part of the difference in both theory and output. In the History section, Scott Rettberg observed that a lot of hypertext fiction tends to be metafiction about the nature of hypertext (see e.g. page 33). There’s metafiction in the IF canon as well, especially in Adam Cadre’s oeuvre — Shrapnel and Endless, Nameless qualify, and 9:05 and The Nemean Lion are both metafictional jokes. But that’s not the majority output by any means.

Koenitz covers a number of previous theoretical approaches to this field, especially some of Nick Montfort’s work, before introducing his own model, which defines a “protostory” as a “space of potential narratives”, including ingredients such as settings, assets, definitions and code. He also suggests “narrative vectors”:

The term narrative vectors describes substructures within a narrative design that provide a specific direction. Narrative vectors work not as isolated structures, but rather in connection with the preceding and the following parts of the narrative. The purpose of such structures is to convey important aspects to the interactor, to prevent an interactor from getting lost and to authors in retaining a level of control…

Narrative vectors in Afternoon are combinations of lexias and links that are designed to create specific experiences, for example the revisiting of a particular passage after the interactor has gathered additional knowledge… Narrative vectors in Façade determine if an actor is kicked out or if she/he reaches the therapy part in which Grace and Trip are able to rescue their marriage. (99-101)

Koenitz argues, ultimately, that this model opens up valuable analysis techniques: “…analysis is no longer constrained by the need to adapt legacy theoretical positions and can instead fully focus on describing the particular narrative strategies of Afternoon and Façade.” (101)

And considering my stance above, my reaction to this is a combination of “well thank goodness for that, we can finally get down to business” and “why do we need a new set of vocabulary to discuss these effects that, in fact, many people outside IDN academia have been talking about in detail for many years?” Do protostory and narrative vector give us leverage on those problems that we didn’t have before? The two examples Koenitz offers appear pretty dissimilar — the underlying code of Façade is a wildly different beast from the underlying code of Afternoon — and it might seem that he means something as broad as “mechanics” or “implementation.”

But elsewhere he also notes that “narrative vectors are roughly functionally equivalent to plot points in legacy media” (100). So perhaps a narrative vector should be understood to mean “those code and design aspects of the piece that control which content will be accessed in what order.” That does specify a useful subset of the whole design of an interactive piece; it might, for instance, exclude HCI questions of whether the player is accessing these elements via tapping or via typing or via gestures read by a webcam; it might also exclude minigame mechanics, if the minigame contains nothing of narrative significance.

The two chapters that follow, “Emotional and Strategic Conceptions of Space in Digital Narratives” by Marie-Laure Ryan and “A Tale of Two Boyfriends” by Janet Murray, address themselves much more to explaining specific phenomena in specific games.

Ryan’s piece proposes a distinction between space (open, undifferentiated) and place (specific, emotionally resonant), and uses this distinction to talk about our experience of open-world games, Second Life-style simulation, and Tale of Tales’ The Path, where important places are embedded sparsely in a much larger open space.

“A Tale of Two Boyfriends,” meanwhile, opens with a critique of Prom Week as a narrative experience:

For example the impressive social simulation Prom Week (McCoy, 2013) offers eighteen multi-parameterised characters whose interactions are controlled by a social physics that is ‘based on a set of over 5,000 sociocultural considerations.’ The highly detailed back end does not necessarily make for more engaging storytelling… it is hard to care about Zach and Monica and the other sixteen characters because they are both under-dramatised and over-specified. The emphasis on simulation over storytelling can create a kind of uncanny valley that is neither game nor fiction, a problem that developers addressed by switching their focus from open-ended storytelling to the setting of goal-directed puzzles in the form of social games. (121)

This certainly accords with my own experiences, both about the experience of playing Prom Week (which I nonetheless enjoyed and found fascinating, but from which I would now be hard pressed to identify even a single significant plot beat) and about the experience of writing narrative work around a social simulation.

However many aspects of personality we were notionally able to model in Versu projects, it was generally most effective from a gameplay perspective to focus on the aspects that mattered to a particular story or indeed a particular scene of that story: Blood and Laurels social practices pertain to threats and perceptions of loyalty, for instance, whereas social practices in the Regency scenarios are more focused on flirtation, commitment, social respectability and status.

This is one of many issues that makes me doubt it is practical to build one single social simulation library (however large and nuanced) that could possibly apply to all the games one might want to build or all the stories one might want to tell.

In place of this kind of social simulation, Murray suggests one built around formalizing the underlying moral scheme of the story being told.

As a case study, she suggests a scheme of story in which the heroine must choose between the Boyfriend of Obligation and the Boyfriend of Desire. The Boyfriend of Obligation is the romantic partner the heroine is “supposed” to be with: her husband if she is married, or the person who is of an appropriate social and financial status if she belongs to a status-focused society. The Boyfriend of Desire is the partner who represents risk, freedom, and the fulfillment of her own wishes. Murray points out examples of this scheme in a variety of stories: Helen/Menelaus/Paris; Guinevere/Arthur/Lancelot; Alicia Florrick/Peter Florrick/Will Gardner in The Good Wife. Although Mr Darcy is ultimately a positive boyfriend of obligation — since he could hardly be more eligible — Elizabeth first has to deal with Mr Collins, definitely an Obligation figure, and Mr Wickham, an unsafe Desire figure. Murray then addresses the archetypal characteristics of women in these scenarios. She proposes an Austenian interactive narrative that makes use of these (and with explicit reference to the Versu Regency stories):

I envision a somewhat similar digital adaptation, further from the particular plot of any existing novel but more closely based on Austenian methods of character abstractions. Such an interactive story would present the interactor with an array of potential suitors, and invite her to use appropriate social rituals to uncover clues to the underlying selfishness, intelligence, and honesty of each of them. The suitors would not be based on any existing characters, but they would differ from one another in ways similar to the ways in which Mr Collins, Mr Bingley and Mr Wickham differ from Mr Darcy. (132)

Reading this, it was hard not to think: “but… dating sims? visual novels?”

It doesn’t have a particularly naturalistic interface, and the relationships are all female-female, but Black Closet more or less is this game, complete with the contrastingly-specified NPCs, the need to determine the loyalty of the other characters, and the application of social strategies to elicit people’s true traits. The protagonist’s choices affect her relationships with other characters as well as her own moral stance, and the effects ramify in quite a lot of potential story branching. It’s perhaps my favorite in its genre, but there are lots of other dating sims that track numerically more or less what Murray describes here: Cinders is another game with a similar strategy, though its mechanics are much simpler; Matches and Matrimony and Regency Love do the dating sim with Austen theming; Hatoful Boyfriend combines the dating sim features with a darker, hidden plot.

Meanwhile, some of the same principles apply (though much more loosely and fluidly) to the design of characters in some tabletop storygames, where NPCs are designed specifically as foils to player characters in particular ways (see for instance the character creation process of Polaris).

This (again) is not to say that Murray necessarily should have encountered all of these or that she would necessarily agree with me that they meet her criteria. People without prior experience in VNs and life sims or dating sims sometimes bounce off Black Closet, I find, because stylization and the mechanics focus them on what is most artificial about the piece, and make it hard for them to access the underlying genius of its construction.

In any case, the idea of a dating sim high-level structure but with a somewhat more natural dialogue exchange does appeal to me. I like the Boyfriend of Obligation and Boyfriend of Desire scheme specifically; and I am  certainly interested in general in the problem of finding the right abstraction layer for narrative simulation, as Murray urges us to do:

As these three distinct but overlapping traditions continue to grow in ambition and expressive power, it is important that we distinguish the aesthetics of storytelling from those of gaming and systems engineering, and that we create computational structures that allows us to focus on the abstraction layer in which traditional narrative creativity takes place. (134)

The last chapter in this section is Nicolas Szilas’ “Reconsidering the Role of AI in Interactive Digital Narrative.” Szilas is one of the researchers behind IDtension, an academic interactive narrative system that models both theories of narrative and the response of the audience. (On that same site is “Nothing for Dinner,” an IDtension game that you can play if you want to get a sense of the system: follow the links in the righthand column. You will need a browser that is compatible with the Unity web plugin, which is to say, not Chrome.)

His chapter talks about the challenge of designing tools for AI-driven interactive narrative, and the need for an iterative process where the tool is prototyped and then repeatedly refined to reflect how actual authors use it and how they conceptualize their work within the system (141-2): I strongly agree with this but have often seen system creators approach their task as though they can expect to complete a tool before it is ever used to build a single story.

About this process, he says:

The creative view of AI fosters the development of a new logic that handle concepts such as action, narrative goals or dramatic conflict directly, not as applications of known AI algorithms. In other words, the computational models become natively narrative… (144)

He argues that it is hard to create highly interactive digital narrative from first principles, and that this kind of research requires long-term interdisciplinary projects that don’t have a natural home in either the game industry or academia; that the most successful interactive narrative projects arise from game designers who float between multiple academic and indie contexts, and thus have the freedom to combine the best of both worlds.

And:

[Because we don’t agree on the measurable qualities of a good narrative] it seems difficult to formulate HIDN in terms of a problem to solve and then apply AI algorithms to solve this problem. In fact, researchers tend to define a different problem and attempt to solve it, which may lead to relevant scientific results but may not serve the purpose of HIDN. Rather, HIDN should be fundamentally formulated as a design activity, aiming at building a series of software projects that satisfy the author and then, consequently, the end-user. (143)

Preach.