Subcutanean releases in February 2020 and can be preordered/crowdfunded at Indiegogo.
Byron Alexander Campbell: Your past work includes twitter bots, interactive fiction, generative texts, collaborative storytelling games, and whatever the hell Ice-Bound and Subcutanean are. How would you introduce yourself to an unfamiliar audience?
Aaron A. Reed: I’m a writer and game designer interested in exploring new ways that creators and players (or readers) can work together to tell stories. I’ve been lucky enough to have been able to pursue this idea in a bunch of different ways: as an art student, as a computer science PhD, as an indie dev, and most recently in the games industry.
BC: Your latest project, Subcutanean, is a print-on-demand horror novel created procedurally, via the magic of computer technology, so that no two copies are alike. There’s an interesting tension here—the reader is confronted with the knowledge that a vast number of other possible texts exist but, unlike in a video game or a Choose Your Own Adventure book, is prohibited from accessing them. How do you hope readers will engage with this premise?
Each time a copy of Subcutanean is printed, a new reality is minted.
AR: The plot of the book involves two characters who discover a gateway to an underground space that multiplies possibility, and the story hinges on the tension of wondering if there are parallel realities where things turned out a different way, or where different possibilities might exist. I thought it would be really interesting to make readers feel the same way as the characters: constantly wondering what other versions might exist, if they’ve got the “best” one for them. It’s a horror novel, but that adds a layer of existential meta-horror, if you will. Like the characters, the readers are unsure what else could have happened.
As an aesthetic experience, this is something that I think happens a lot in digital storygames: wondering if you made the right choice, if you’re down the best branch. And back when I was growing up, there would be a lot of urban legends around certain games: whether certain characters would inevitably die or if there was some way to save them, whether you’d found all the secret areas, etc… I thought it would be interesting to try to bring some of those conversations to the experience of reading a printed book. If you and a friend both play a game and talk about it, you’ll be sharing stories about your different experiences—so what’s it like to have a conversation about a book when no two readers got exactly the same story?
BC: I remember those rumors. It seems like some of the magic and uncertainty that accompanied these types of experiences has been lost as the Internet has matured.
AR: It’s different when there’s an exhaustively researched wiki for everything! I thought a lot about whether I wanted to restrict access to the master text of Subcutanean, to try to preserve that sense of mystery—but ultimately figured there would be so many people curious to see behind the curtain that in the end I did decide to make it a thing you could get if you wanted (there’s a higher pledge level with a special USB key containing all my source code and the master text).
Subcutanean is a novel of impossible spaces.
BC: This is also a queer narrative; you’ve written that Subcutanean was inspired, in part, by your experience of coming out in a conservative state. Can you expand on how that interplays with the speculative/horror elements of the story?
AR: Something that good horror does incredibly well is externalize our internal fears about ourselves or the people we love. Some of my favorite horror movies this decade, for instance, are masterful in how they do this: The Babadook, Get Out, and It Follows are some that spring to mind. In It Follows, for instance, the relatable teenage fear that losing your virginity will somehow taint you, become a dark mark that will follow you around forever, inescapable, is translated so neatly into an external monster that it works even you aren’t consciously aware of this subtext. The terror hits you on a deeper level, and feels more true—and more terrifying—because of it.
In my own awkward and delayed coming-of-age process, I remember how painful it was to always feel like I was doing everything wrong, wishing there was some other version of me that had figured himself out, was more confident, knew how to get what he wanted out of life. The way that becomes externalized in this book into a huge and ominous physical place that the characters get lost in grew out of those anxieties and fears very organically. Representation has been important in a lot of my past work, but this is the first time I’ve really based a piece entirely around that time in my life, I think because it took me a long time to get enough distance and perspective on it to understand what I went through and how I grew up enough to get past it.
Blue Lacuna is a longform work of lyrical interactive fiction.
BC: On the subject of queer representation, I remember that Blue Lacuna opens itself up to that rather organically, one of the many things about it that felt cutting-edge in 2009. As a genre, storygames and interactive literature in particular seem to have embraced this style of queer representation; Fallen London, for example, was (I believe) a pioneer in presenting nonbinary character creation options and putting absolutely no restrictions on who (or what) your avatar might choose to romance. But that kind of commentary-free inclusion, while important, isn’t the entirety of representation.
Speaking of Blue Lacuna, that was my first encounter with your work: a pay-what-you-wish interactive fiction novel in the tradition of old-school text adventures like Zork and Adventure (but in many ways, as I mentioned, cutting-edge—Blue Lacuna is one of a handful of IF works I will consistently recommend to people who have no history with IF). How has your background in interactive fiction and interactive storytelling informed what you’re doing now with bots and generative text?
AR: I kind of see what I’ve been doing over my whole career as a huge trip around the boundaries of what an interactive story can be. So Blue Lacuna and my other text adventures are a very simulationist approach to being part of a story: there’s a whole little world that’s simulated and that the player moves through, one action at a time, and part of that game’s project was to ramp up the details of the simulation (of both setting and character) to make something that felt truly responsive to the player. I did a piece called 18 Cadence some years back that put the player in a more editorial role: they’re given all these pieces of a hundred-year history and asked to try to assemble them into a meaningful story, without being able to act directly or change anything about it themselves.
More recently I’ve been fascinated by indie tabletop storytelling games, which are coming up with all kinds of interesting rules and mechanics for helping untrained players collaboratively improvise a compelling story (my game Archives of the Sky was in this tradition). And Subcutanean, while not a game, is still a part of this same overarching experiment: what does it feel like to have procedural text with no interactivity at all? What can we learn about using procedural text in other places (like games) from that experiment?
The text can vary in large and small ways.
BC: When, why, and how did you get started working with procedural text?
AR: I’ve always been fascinated by it—somewhere in a crate I’ve got a hand-written and illustrated Interplanetary Spy book (an off-brand Choose Your Own Adventure series I was super into) written at around age 8 or 9. I taught myself coding at a pretty young age primarily to make my own text adventures. As an adult, I’ve always been intrigued by the way text can be seamlessly modified or customized to create something that looks hand-written but can change for different readers. My text adventure Blue Lacuna has tons of fiddly code for writing room descriptions that can vary based on the time of day—the beach at sunrise gets described differently from the beach at noon, or sunset, or twilight, or midnight—and I love that players don’t especially notice that: it just seems like the way the text should be for that particular moment.
BC: So do you view procedural or generative text and storygames/interactive fiction to be essentially the same thing? To me, there’s a sharp distinction.
AR: That’s an interesting point. I think as someone who’s always come to games from the perspective of a writer, procedural text has always seemed to me the most promising tool for making interactive fiction truly interactive. In my dissertation, I talked about the paired notions in game studies of “expressive input” and “expressive processes”: the capability of a game to, respectively, understand or execute complex ideas that convey a unique meaning. So SimCity, for instance, has both expressive processes that convey Will Wright’s unique ideas about city planning and expressive input that lets players build their own idiosyncratic cities. But it’s still rare to see games offer that kind of expressivity in their actual stories. Systems for assembling and authoring procedural text are a way of breaking down the current building blocks (a fifteen-second MP3 file of a character speaking, for instance) into smaller units (phrases and words); once you’ve done that, you can start combining them in new and surprising ways, and that opens up a door to whole new kinds of gameplay.
There’s obviously a whole tradition of generative text outside the context of games, of course, but they’ve always been married pretty closely in my own work, even Subcutanean.
BC: This project deals with quantum possibilities and parallel realities. Any inspiration from Bandersnatch?
AR: Subcutanean was actually mostly written back in 2017, so no direct inspirations! But I think it’s only natural when you’re writing in a format that allows multiple suspended possibilities to start thinking about fictional premises that involve those, too. I also think it’s really interesting that we’ve had the technical capabilities for dynamic text and story for a long time now, but are only just starting to see them impact the mainstream in a serious way. I think part of that is that it’s such a fundamental paradigm shift for storytellers that it really did take waiting for a generation who grew up playing interactive narratives to start to understand what they’re good for artistically and how to use them to tell fundamentally different kinds of stories.
Two versions of page 112.
BC: Do you think procedural generation is having a moment?
AR: I think it’s a really weird and fascinating time. In 2014 I co-created a game, The Ice-Bound Concordance, that involved a dead writer brought back to life as an AI. Now, five years later, with tech like GPT-2 that can plausibly generate text in certain styles, that seems way less like science fiction. I would not be at all surprised if sometime in the next five years we start to see companies like the ones in Ice-Bound that offer endless stories written in the style of famous authors like J. K. Rowling or George R. R. Martin.
I think there’s a real tension between procedural generation as a tool for writers (or artists or musicians or whoever) versus trying to use it to replace them. So with GPT-2, for instance, I’m much less interested in reading the output of your GPT-2–trained bot than I am in seeing tools emerge that let creators collaborate with that tech to bring new kinds of experiences to life. The real question for me is how we as humans can harness these techniques to continue to share the experience of being human with each other.
BC: That’s very much what Ice-Bound was about for me. One of the things that I found so fascinating about that “game,” for lack of a better word, was how ungamelike its interactions were. As the player/reader, I didn’t get the feeling I “was” any of the characters; instead, I was acting more as an author or editor, choosing between small details to affect the tone of the unfolding narrative. I think the idea of choice and branching narrative in a lot of story-driven games has gotten kind of polluted by mechanical considerations, to the point where (in mainstream, AAA games especially) I often find myself making a decision not because it’s right for the character or the story but just because it will yield a quest reward that suits my playstyle (this was a huge problem in Fallout 3) or push me further along in one direction along an ultimately binary morality track (as in Infamous or Mass Effect). So it was refreshing and oddly liberating when Ice-Bound stripped those mechanical considerations away. Subcutanean, by removing interaction altogether, seems like an extension of that experiment in some way.
Like Subcutanean, Ice-Bound’s shifting text varies in ways that are primarily tonal.
AR: I’m glad that aspect of Ice-Bound clicked for you, and yes, this new project is playing with some of the same ideas. In Ice-Bound you don’t control characters directly, just make high-level decisions about what kinds of themes and plot elements you think should be part of the story. Here you don’t even have that kind of control, and yet you’re still reading the story with this frame of “Why are things playing out this way? What else could have happened?” I’m especially curious to be a fly on the wall when people start having discussions about their Subcutaneans together, advocating for or against certain versions as being better or worse. It’ll be a bit like transplanting some of those Ice-Bound conversations [with the AI, Kris] into real life.
BC: From reading your (excellent) design posts on Medium, I get the impression that the crux of Subcutanean’s variability comes from, well, variables that can be turned on or off to affect the way scenes, chapters, or the entire book play out. The narrator could be terse or verbose, drunk or sober, boiling pasta or stargazing. Roughly how many of these variables are there in Subcutanean?
AR: There are hundreds of places text can randomly vary, but the variables help make that selection feel more intentional. So for instance, one of them decides which character from the narrator’s past will appear in a pivotal scene towards the middle of the book. Depending on which character is chosen, that triggers different setup scenes in earlier chapters to establish that relationship and foreshadow the coming confrontation.
So there are about a dozen of these “major” variables that each touch multiple bits of plot across the whole book. There are maybe another 25–30 smaller variables that only touch specific chapters, controlling things like which of several scenes might happen at a particular moment, the details of its setting, whether it’s played up into a big moment or moved past more quickly, and so on. And then there are a handful of less significant ones, like who the narrator’s favorite author is, that are consistent across multiple scenes but pretty minor in their impact.
You mentioned “terse vs. verbose,” and that’s the other major category of variable: not manually applied, but factored by the rendering algorithm into how all the smaller alternatives are prioritized. So a verbose narrator will prefer a longer way of saying something than a shorter one, but since these all work together, a verbose narrator who also prefers alliteration and hates the passive voice might go with the shorter version if it ticks both of those other boxes.
Figuring out an interesting set of “narrator variables” was actually a really informative process, because sometimes it would be obvious that one of the narrators was clearly superior to the other: active vs. passive voice is actually a good example of that. But because I could click a button and say, “Show me all the variant text that uses the passive voice,” it made it easier to find places where that was just weak writing and eliminate them, eventually retiring that particular narrator as no longer useful. It was a really interesting way to go about the process of editing and refining the text.
Subcutanean uses a specially designed syntax to enable variability without unnecessarily interrupting the flow of writing.
BC: It’s fascinating to read about how you used these narrator variables as a revision tool—the way you describe it is almost Darwinian. You also wrote this in one of your design posts: “When revising and rewriting a list of alternatives, I knew I was done when I could no longer decide which one to mark as ‘best’: when I’d be equally proud to have any of them carry the story.”
Both of these recent projects, Ice-Bound and Subcutanean, appeal to me as a writer because they suggest an alternative to the anxiety-drenched process of revision. I have this constant fear that by altering or deleting something in the text, I’m “losing” something raw and important. So it’s conceptually attractive to have this alternative approach ti writing where revising one scene doesn’t have to erase the previous version. Was this a consideration at all when conceptualizing Subcutanean—that is, do all of these alternative versions and variables exist, in part, as a way to avoid the horrible work of “killing one’s darlings” and propping up one version of the text as “final” and “definitive”?
AR: It can definitely be a double-edged sword! When revising, I would very frequently get into the exact pattern you’re predicting, of adding an alternate version of something instead of replacing it, in cases where I wasn’t sure if the new one was actually better. This definitely felt a lot safer, and there were certainly occasions where I’d come back and realize the new version was in fact actually worse: I probably edit my own stuff too much for its own good, and sometimes lose the spontaneity of early drafts that way.
But you’re absolutely right in that it can be a crutch, too. At the end of the day, you still have to be an editor, deleting the stuff that doesn’t work and relentlessly revising the stuff that does to make it better. There are a lot of passages in Subcutanean that started out one way, went through a draft with three or four variations, and ended up back with no variations again, because it was clear that one way I’d written it was the only right one for the story. Even in a “story possibility space,” there’s still a lot of possible stories or ways of telling them that are wrong. Writing this way lets the hedge grow wilder than usual for a while, but you still have to prune it back down again in the end.
Four ways into the novel.
BC: Because of the peculiar needs of this project, you ended up creating your own procedural text authoring tool in Python specifically for this book. Reading about this in the design posts, I marveled at its elegance. Do you have any plans to make the .quant format available as an authoring tool for other writers?
AR: I would love to find the time and space to do that, and I’ve actually put that as a “stretch goal” in the Subcutanean crowdfunding campaign. The .quant format is interesting because it’s focused on functional minimalism: it’s very simple from a technical perspective, so all the design work went into the question of “How do you create an authoring format for writing variable text that stays out of the author’s way as much as possible?” So much of my past creative work in generative text has been unduly stressful because I was having to juggle programming or syntax details in my head, leaving less cognitive bandwidth for the actual writing. The .quant format was really an attempt to strip away as much of those technical stumbling blocks as possible.
BC: Do you worry that any serious literary ambitions you have with Subcutanean will be eclipsed by its central gimmick? That is, do you worry that people will talk only about the fact that every copy is a unique text and it uses all of these generative tools in its creation and overlook the story itself? Or is that, in some way, a desired outcome?
AR: I think I’ve worried about that for all of my projects, actually! I did believe enough in early drafts of Subcutanean that I considered trying to find an agent and release it as a traditional novel. But now that I’m almost finished with the multilinear version, I just can’t imagine this particular story being told any other way: it’s so obsessed with alternate possibilities and subtly different versions, it feels right for that to spill out into the format itself.
BC: Besides Subcutanean, what are you most looking forward to—either another project you have in incubation or something you’re waiting to see, read, or try?
AR: I’m super excited for the Lovecraft Country series that Jordan Peele is producing. It’s been really fascinating to see in the last decade how horror creators have started to make art that honestly engages with both the good and bad parts of his [Lovecraft’s] legacy—I also recently read Paul LaFarge’s The Night Ocean, in that same vein—and a longform format gives so many great opportunities to engage with those ideas in a more sustained, ongoing way than a book or film.
BC: Thank you so much for taking the time to answer these questions. I’m excited to see the impact Subcutanean will have.
Aaron A. Reed is a writer, designer, and researcher focused on finding new ways for gamemakers and players to tell stories together. His fiction, games, and playable artworks have won recognition from a broad range of storytelling communities, including the Independent Games Festival (video games), the ENnie Awards (tabletop roleplaying), and Kirkus Reviews (traditional publishing). Aaron is a multi-time IndieCade and IGF finalist, and his work has also been shown at South by Southwest, Slamdance, and GaymerX; he has spoken about digital storytelling at PAX and PAX East, Google, WorldCon, NarraScope, and the Game Developer Conference. Aaron holds a PhD in Computer Science and an MFA in Digital Arts and New Media. He lives in Santa Cruz, California.
