Iterating San Tilapian Studies

Over the years since I first ran it, I’ve made a few iterations of my narrative (large-)party game San Tilapian Studies. It’s also been run by other people with their own versions of the stickers and text, and even been made available to the public at the Wellcome Play Spectacular.

In 2015, I offered a version of it as an IF Comp prize, and the winner requested that I adapt it to a particular fantasy world idea he had in mind. That version was mechanically pretty similar to the original — partygoers are given stickers with nouns or descriptive phrases on them, and must find other participants that are a good match for them in order to make a three-segment-long description of an object. Then they make up some backstory about that object and write it in the book.

The main tweaks I made were:

• the game package also included a fantasy outline map that players could use to locate places in their game world
• the sticker collection included a few wild-card words, words that could be played as either a noun or an adjective (like “bone” or “mirror”). The intent there was to help keep things moving smoothly if people felt they were running out of plausible matches

This year I offered the San Tilapian Studies kit again. One of the winners requested the original version, which is cool — it’s fairly easy for me to do that, and I think the Ruritanian romance setting provides a lot of room to imagine different miniature plot lines: love stories, politics, military affairs, interpersonal intrigue, etc., all fit into that world.

The other winner asked for a Lovecraftian re-skin of the set, and that was a bigger revision.

The rest of this post is about the design work I did to adapt San Tilapian Studies, first for a minor upgrade and second for a larger re-skin. While the result isn’t played on a computer, this is an exercise in developing a corpus for procedural use.

Designing, testing, and editing a corpus is in fact a significant part of the work in any procedural text project — so while this is a bit different from, say, Annals of the Parrigues in that it doesn’t result in a finished single text, a lot of related issues come into the corpus design.

So, first the upgrade of the original setting.

I am not quite using the original sticker set purely because when I originally wrote the game, the grouping included a reference to a “painted gypsy caravan.” This was raw ignorance on my part. It may be relevant that I was still living in the US at the time, where popular awareness about that word is I think a bit behind where it is in the UK; at any rate, I didn’t at the time know that the term was used as a slur. Now I do know, so I’ve replaced that sticker with something else.

Less importantly, I’ve swapped out a few other stickers, mostly for three reasons:

• Setting consistency. The original included “a skull with three eye sockets”, which is visually evocative, but suggests a supernatural or science-fictional cast to the story that isn’t really borne out by the rest of the set.
• Political themes. Your basic Ruritanian romance is nostalgic for a time of royal families and aristocratic flourishes, and in the 5+ years since I made the first set, I’ve become more cautious about nationalism and about equating inherited wealth/status with being a good person. So I added some elements that subverted that a little.
• Strengthening the “function” stickers. During collaborative storytelling, players often feel they need permission to commit to drastic outcomes — and the created story is usually better when they do. So I took out stickers that said things like “reflected badly on someone” or “caused indigestion” and added ones like “ruined a reputation” and “prevented a coup.” It would be totally possible to take a story sequence about indigestion and turn it into something more politically significant — but offering the players high stakes most of the time seemed like it was more likely to generate some sparky content.

So that was a comparatively light upgrade pass.

So, about Lovecraft. This project makes the second time in the recent past that I’ve done a procedural text project on a Lovecraft theme: my Cragne Manor room had already had me looking up commonly-used Lovecraftian words, which have conveniently already been studied by lots of other people, so it’s not as though I needed to calculate anything myself there.

For the San Tilapian Studies Lovecraft set, I initially experimented with pulling in a bunch of generic objects and Lovecraftian adjectives, and generating the text that resulted.

(Sidebar: for the text generation in this project, I used Spirit AI‘s Character Engine authoring tool because, well, it’s conveniently to hand, and it’s faster to set stuff up there than my previous generator in Inform. Character Engine also features some word-suggestion tools to give you proposed variants, though I used that sparingly on this project. In practice, one could also use Tracery or something else — I wasn’t really leveraging any of the other Character Engine features that set it apart, and this is about the simplest generative grammar imaginable.

Whatever one uses, I find one does want some computer-aided method to test the components for a game like this. It’s a fast way to figure out whether the generated combinations are looking boring, or whether there are some sticker ideas that seem cool in theory but in practice just never quite work.)

In this case, that first pass was clearly not that satisfying. It looked a bit like this (and I’m reconstructing a bit after the fact because I only decided to write this up later in the process):

Procedural text generation is a great vehicle for parody, because it points out the typical habits and tropes of the thing it’s picking on. The problem is, this time, I wasn’t setting out to do Lovecraft parody. I wanted to give players the components of a story prompt that would spark discussion and creativity, not one that would remind them of Lovecraft’s most repetitive aspects. And I wasn’t helped by the fact that a lot of Lovecraft’s most-used words are on the vague side. I’d guess a lot of people would define “squamous” as “a Lovecraft word for something that’s somehow icky,” and it’s hard to specifically picture what qualifies something as eldritch.

The other, subtler point here: all of the above combinations would basically work, if you were willing to imagine a letter or manuscript of unusual size or material; whereas what you want for San Tilapian Studies is a grammar in which some combinations are clear mismatches, but any given sticker could probably work with half to two thirds of the other stickers. It’s the lumpiness of distribution that makes it fun for people to go around comparing stickers until they find something good.

So I scrapped that first word set and had a different go with more specific terms. This forced me to actually go back and read some actual Lovecraft, rather than any of the numerous derivatives. This also made me realize that there were thematic points I could be picking up on along with the vocabulary. (And some I didn’t want to pick up on, of course. “The Shadow over Innsmouth” is not just sort of incidentally racist; it goes into deep and extensive detail about the protagonist’s fear of anyone who doesn’t look 100% Anglo-Saxon, and the whole story is motivated by a horror of “gently bred” people having children with The Other Sort.)

So I moved towards longer phrases rather than single adjectives. Lovecraft’s archaism and ornate sentence structure also influenced me — I tend to soak up some qualities of what I’m reading — and this sticker set came out very much wordier than the fantasy one from a few years ago.

(This screen is only partway through the process; eventually I ditched most of the words like “intolerable” entirely, and neither the trench coat nor the dagger made it through to the final round either.)

As when I was building the original San Tilapian Studies set, I made word/phrase lists that were much longer than I needed, then did a lot of procedural rolls, then cut or rephrased anything that didn’t seem to be workable a good percentage of the time.

Another thing that struck me, reading a couple of the stories, is how much Lovecraft is playing with subjectivity, and with people not being sure whether they’ve really seen what they think they’ve seen; or not being sure they’ve understood what they saw because it was so mind-boggling. That led me to introduce quite a number of “function” stickers that were in the first person: “worsened my husband’s condition”, say, or “revealed a pattern I had missed.”

(It struck me that I am not sure Lovecraft would write the kind of protagonist who might have a husband, but if not, it’s another deviation from his worldview that I’m happy to make.)

Meanwhile, there are some stickers that describe not just sinister evidence, but Actual Horrors — monstrous winged creatures, gelatinous sea-beasts, the usual — but the rest of the combination would determine whether that item was really present, pictured/reported only, or indeed dreamed or imagined. So a player’s “black winged creature” might match to create “a black winged creature sketched on the flyleaf of a guidebook presaged worse things to come” or “a black winged creature moving at terrible speed pursued me in my dreams” or “a black winged creature from the planet Yuggoth obliterated Kingsport.”

Quite possibly I’m overthinking this, but I really like that these sticker combos mean that the reality-level of objects can only be ascertained through confirmation by other players.

The final set also does achieve the lumpy distribution effect I was looking for. The midnight bus probably wasn’t freed by an earthquake or suggest a monstrous ancestry, unless we came up with a really elaborate story to make those bits fit together. But the photograph of the old woman’s face is easy to explain; the corpse and the tabloid paper, only a bit trickier:

Finally, I’ve added a little extra optional thing to the game, if the recipient chooses to deploy it. In the sessions of the game that I’ve run, there have generally been some people who felt “done” after one or two matches in the book, and then a handful who were super into it and kind of wanted to keep playing in some manner after they’d used up their stickers.

So in the Lovecraft set, I’ve added a small set of stickers in a fourth color. The gameplay still proceeds with players matching in threes, but if after an original couple of matches, people are really enjoying themselves and want to keep engaging with the game in some fashion, they can then draw a fourth, yellow sticker, which might say something like “Lies!” or “I was there”:

These are meant for the player to use on someone’s already-written page (because by this point in play there should already be a number) and add some extra content about that same object, only from a different narrative viewpoint.

(In Parrigues Venom/Salt/Mushroom/Beeswax/Egg terminology, this is adding an extra Egg aspect to the game.)

I liked the idea of giving interested players a way to stay engaged, so I went back and added some scholarly-footnote options to the classic San Tilapia set as well — but themed more around historical interpretation than around personal subjectivity.

Next up: sending the stickers to the printer.

Elsewhere:

• Fiasco playsets are a different take on capturing tropes from something. (For instance, here’s a Lovecraftian Fiasco playset.)
• Fiascomatic is Ian Horswill’s framework for automating Fiasco playset generation.
• Rewordable is a case of doing some text frequency studies in detail in order to create an awesome card game play experience