Procedural Storytelling in Game Design

edited by Tanya X. Short & Tarn Adams

I think I need to get this posted before graphical Dwarf Fortress releases and all the proc gen people disappear into their steam decks forevermore. Truly I have no idea how interesting this book might be to people that haven’t spent years of amateur dedication frolicking in the intersection of narrative design and video games already. I hate to start off with a “good for fans of the genre” consumer report, but this is a book that some might find straight-up useful, rather than fun to read. I, however, found it fun to read because my brain is filled with deep lore about chatscript branching, ELIZA, Galatea, plump helmets, and the Amulet of Yendor. The ability to recognize any or all of those references may push you towards Procedural Storytelling; if you’re anything like me, you will somehow recognize nearly every proper name in the book, either from reading something else or from playing the games listed. Familiarity with a game title or company name, a cited designer or essay writer likely gave me blind spots to the poignancy or cohesive feel of the essays; to me, nothing was that hard to follow, but if I didn’t know what Sunless Sea was or had never wishlisted Overland before, I’m not completely sure if there are enough broad strokes to make reading about how they use procedural generation compelling. I simply cannot say how much a priori experience is needed to follow along!

Because I personally believe that procedurally generated narrative emergence is the future of game design, reading about where the space is now* by some of the most interesting and experienced creators out there is a wonderful experience (*as of 2019).

If this thrills you, you might like PSiGD

Example: The very same morning I finished the (excellent) Tarot as Procedural Storytelling chapter by Cat Manning she was on a podcast I was listening to about Crusader Kings 3. This is the type of emergent narrative upon which proc gen relies—creating an ecosystem where these little coincidences might happen and letting the player (or general life experiencer, in my case) weave together the threads—and kismet like this allows any series of random events to feel important or special. Across the whole wide world, experiencing Cat Manning twice in one day seemed cosmic: now imagine a closed system wherein someone can create those singular moments with self-reinforcing contained algorithms and the smaller worlds of a proc gen gamescape. Those co-incidents (as opposed to coincidences) will proliferate, and their interactions will make the game shine.

Most of the chapters deal specifically with a particular video game—the chapter on tarot reached far outside of the standard electronic systems to make its point, but most focus on a procedurally generated game and how that system works to enhance or support the gameplay loop or narrative. Each chapter and each writer is unique within that structure. But even something as far afield as mystical fortunetelling cards has grounding for me as a specific video game callback; I first experienced tarot decks through the SNES game Ogre Battle: March of the Black Queen, which featured them as gameplay mechanic as well as part of the character generator and general aesthetic. While Ogre Battle was not procedurally generated, it carried the same spirit as proc gen: “set the parameters and see what the system does.” Toss in a dash of “I cannot quite intuit what is going on under the hood of this game” and you get the exact thing I love—a game where your actions and choices constantly return unexpected discoveries!

Tarot is a neat storytelling device (someone else thought so too, as whoever had my library book before me marked up this, and only this, chapter), but I never gave it too much thought as a narrative device: Tarot as Procedural Storytelling created an excellent mental space to ponder the whys and wherefores of mechanical randomness and cohesive narrative flow. Thanks, Cat Manning!

You can always count on reddit to have what you need

And while I could (and might?) go through and thank each and every person that wrote or edited this book, there are a few key takeaways that are too important not to surface. Foundational understanding begins with Kate Compton’s 10,000 Bowls of Oatmeal problem, which is cited and referenced dozens of times throughout a number of chapters:

“I can easily generate 10,000 bowls of plain oatmeal, with each oat being in a different position and different orientation; mathematically speaking, they will all be completely unique. But the user will likely just see a lot of oatmeal. Perceptual uniqueness is the real metric, and it’s darn tough.” (Dr. Kate Compton, Getting Started with Generators)

Structurally, that’s probably the single most important thing to pull from the whole text. You can generate and generate and generate, but if they’re all technically different without feeling different, no one is going to care. Most of the rest of the book is about technique or preference, which is neat and fills out proc gen in real way, but knowing the root issue and being able to recognize the shorthand when someone mentions the “10,000 bowls of oatmeal” problem is not just helpful, it is required.

The next most obvious of the non-intuitive realizations was how a standard player is only going to see maybe a fraction of your generated output: “Procedural generation is, effectively, a way of getting 200% of the content with 400% of the work” (Bruno Dias, Procedural Descriptions in Voyageur). Maybe make sure that you’re mentally ok with most players not seeing most of your content before you start a proc gen project. (Write a book review website, you’ll learn to live with your like, 90 monthly readers (for which I am eternally grateful <3).)

So if the work is tough and output is unreliable and unwieldy, what is it that draws people to proc gen in the first place? I can only speak as a player on the receiving end of the system, but the large possibility space that procedural generation can output is what draws me to video games en masse, not just proc gen ones: “If player choices don’t change the world, failure becomes a form of “save and reload” or “try, try again.” This causes players to lose their sense of performance, stop role playing, and start grinding out tasks with predictable outcomes.” (Daniel Kline, Dramatic Play in The Sims). Knowing the system can go in a bunch of different ways makes it interesting and worthwhile. If there’s a way I can talk to someone else that played the same game and say, “Wow, that’s cool, that isn’t what happened to me in my game,” then I am interested.

I don’t particularly care about scores, and I’ve stopped really looking for fulfilling narratives; what I care about is agency-derived impact. Actually, it’s even less than that–I care about player input.

Let me back up and use two examples from extraordinarily non-proc gen games—Tales of Arise and Scarlet Nexus—games where the player has no say in anything that happens beyond how quickly you make it to a cutscene. They are both console action/RPGs: their possibility spaces are tiny. You can choose what attacks to use, but they all do the same thing: damage enemies so that they disappear and you can hit the next story scene. What attacks you use really only matter for efficiency—how fast can I get through this random battle?—or role playing purposes—manually disabling Shionne’s healing in battle and not using it until you decide she has become more emotionally attached to her teammates. That is something you could do if you wanted to, but the game won’t react to it, so any impact on the narrative is entirely in your mind. Battles are designed for “grinding out tasks with predictable outcomes,” repetitious efficiency rather than that type of expressive play. Players “lose their sense of performance”—the first time Shionne heals someone, the game won’t know she’s been capable of doing it but holding back for the player’s internal narrative reasons, and there is no real gameplay or narrative reward for doing so.

To be able to feel any sort of player input in games that are expressly linear, I often look for simple cosmetic selections: costume changes or accessories. Being able to maintain the consistency of personalized appearance is often enough for me in JRPGs–if I equip something stupid on Rinwell the mage, for instance, and then she’s not wearing it because I’m being presented with an animated cutscene, well, that’s not my Rinwell. On the other hand, even when she is doing specific actions or pushing forward in a series of funneled plot points or voicing dialogue that I have no control over, if she’s still wearing the swirly glasses that I stuck on her for headcannon backstory reasons, then I feel okay about the consistency of my own (relatively minuscule) emergent narrative. If not, then I’m thinking about why she took off her lucky swirly glasses just to stick them back on the minute the cutscene ended. Don’t they help her to focus extra hard to maintain clear vision and thus increase her magical abilities? Yes they do. In my mind. And why is Alphin in blue again, rather than the mourning red he donned when Shionne was captured? Cutscene Alphin is fake Alphin.

fake alphin is fake

Shiden in Scarlet Nexus started wearing bunny headphones because he’s secretly impressed with Arashi, who always has them on. And you know what? They look stupid. But they are in every single image so far, including the small codec-like talking heads and serious character bonding moments. I appreciate that. Kyoka jumped palettes when we were up in Togetsu, and the huge green lucky charm I had equipped was duplicated when someone left my party and then game back later, but Scarlet Nexus still feels like it takes my cosmetic choices seriously. Both Tales of Arise and Scarlet Nexus are incredibly not procedurally generated games, falling on the exact opposite end of game design; completely linear narratives. The lack of agency writ large works as a great controlled environment for feeling out how even a tiny bit of control can let me as a player feel empowered, even when they I am in no position to build out my own story.

But they also showcase how little I would care to watch someone else play through either of these games, or talk about them afterward. Nothing unexpected is going to happen, no differences between someone on youtube’s playthrough and mine, no narrative purchase with which to talk about “what happened to me” versus “what happened to you.” So the mightiest power of proc gen is expressive play which allows for community narratives—I don’t think anyone is going to read my twitter thread on how Alphin relied on aerial attacks for most of the non-boss battles because they were the most efficient, but I have read hundreds of thousands of words about people creating fortress communities in Kenshi or space colonies in RimWorld (if twitter dies I will miss the threads from IronSpike most of all). This is probably the best way to distinguish between plot and narrative design. “For example, Tetris is a plot generator. “An L block fell. Then a T block fell. It fell in a bad spot. Then a square fell. It had been a long time since an I bock appeared.” (Plot Generators, Adam Saltsman) An actual exception to making tedious plot-based playthroughs of long linear RPGs interesting are text- and screenshot-based Let’s Plays, some of which give arbitrary and completely fictional personalities to the often bland protagonists. Those are, again, storytelling devices, rather than simply walkthroughs: I bet there’s a fun tetris LP out there, somewhere.

The other huge strength of procedural generation, to me, is that games with “unlimited” potentials create too many possibility spaces for the designers to decide what is best. I do not like spending my time in a video game trying to guess what the designer wants me to do. That can sometimes be inevitable, but it still frustrates me when I start seeing the various gameplay inputs as tiny puzzles with a golden-path narrative outputs (see: this review where I complain about using walkthroughs to get the ‘best’ ending in Witcher 3—why even play a game that allows input into the story if you’re going to reload until you get the best one? ARG!). I am not immune to this; when my own brain starts telling me to look for the optimal route so I’m rewarded with seamless narrative breadcrumbing and cohesively designed narrative, I know I am not treating the emergent story like my story but like a book that I have to tape back together to read. Like, Elden Ring is great. I instantly sunk most of my 2022 gaming time into it. But I did what I always do in that kind of game–go in one direction for far too long, snap a couple of quest chains, hear about things out of order, and make the story either irrelevant or a burden. All I wanted to do was help Ranni the Witch or fail in the attempt! Why block her off until I progress other unrelated stuff just for the sake of your story? I can already get to where I think I need to be, I have what she will eventually want, but I haven’t triggered the storyline flag that makes her wake up. Blah.

Procedural storytelling avoids a lot of this because there is no bespoke optimal route–the rise of the roguelike and (to a lesser extent roguelite) genre breaks from prebuilt railroading that restrictive designed storytelling needs to employ. If history can be believed, Rogue’s designer wanted to be surprised by his own game so he could play it, and so had to procedurally generate every bit lest he knew when to expect an orc ambush. So in a proc gen game I don’t look for the “right” way to approach the content, and when things come together in a way that makes little or no sense I don’t blame myself for going out of order. I’m not trying to construct the “best” narrative, because I cannot. This goes hand-in-hand with why twitterbots can throw a bunch of garbage out and still be beloved, and the same reason people (i.e. me) got mad at the twitterbot horse_ebooks–they were curated by a human and thus deceitful in their poignancy. “A well-functioning bot, tended by human hands, is hard to think of as anything but the creation of a person.” (George Buckenham, Things You Can Do with Twitterbots)

So a hand-curated proc gen environment would be too much like horse_ebooks; fake. To circle back to the tarot example, could you imagine your fortuneteller pulling your final card, looking at it, deciding it didn’t make sense, and rifling through the deck until they found one that “worked?” Gross. Not just gross, but pointless. Putting together a reading (aka a narrative) from the cards you are actually dealt is entire point of tarot, and dealing with how random events pop up is the best part of playing video games. It gets to the heart the other key term in Procedural Storytelling, failing forward, “…the idea that when players fail at something, it should advance the story, rather than stop it.” (Steven Lumpkin, Procedural Storytelling in Dungeons & Dragons) And this is the promise of proc gen gaming! No more plot armor for vital NPC quest givers, no more being funneled down the same path after choosing in which order to tackle the first three scenarios (the old bioware Baldur’s Gate gate-and-hub design); a bunch of stuff happens, you deal with it the best you can, and you move on.

I recognize that the reason things like D&D and tarot can work like this because a real human person can improvise, and a standard computer program has to have everything it can do established beforehand; most video games are therefore proscribed in their actions and non-existent in their actual responsiveness:

…because we know what they really mean by “a new gaming experience [every time you play]”. It’s a euphemism, like a lot of language about procedural generation is: “limitless gameplay”, “infinite replayability”, “endless variety”. These are all phrases we see used to describe algorithms that are usually doing the computer equivalent of shuffling a deck of cards and dealing you a new hand. Over time we’ve grown so used to these turns of phrase that we probably don’t think much about the language any more–we instinctively know what the person speaking means, so we don’t think about how they sound to people less familiar with these ideas. (Michael Cook, Ethical Procedural Generation)

Skyrim radiant quests are, like, bad proc gen. Well, the proc gen isn’t bad, but the implementation absolutely is, content without form at all. It is clear from this book that many designers recognize that viewing proc gen as a quick way to “shuffle a bunch of cards” is counter-productive: first, you have to make all the cards; second, most people won’t see most of the cards; third, the way they come together might not make sense or might feel too curated to create anything exciting. So many pitfalls!

Procedural Storytelling in Game Design creates space to think about what people like in video games. A picture forms that encompasses why things like LPs and twitch are popular—people like to talk about “their” game, and to make that interesting the outcomes and minute-to-minute events can’t be predetermined experiences. The strength of storytelling in games doesn’t come from the story the game tells the player, but the story a player creates—and can share with other players—from the tools the game offers up. Without thoughtful systems-level interactions, people playing proc gen games will be doomed to eat unlimited oatmeal, forever.