Random Tables, Inference, & Cognitive Load

29 Mar, 2024

Hot take: the art of using random tables effectively is under-discussed relative to their ubiquity. The following is probably an ill-advised attempt at intellectualizing creativity, but I'm still going to try to articulate a couple lessons I've learned through trial & error.

These notes are informed by my preference for impartial, sandbox-y, 'blorb-y' gaming. I'm not trying to create a grand unified theory of all random tables, just observing what seems to work best for me. Also, while I'm sure there are important distinctions between using tables in prep vs. in play, for now I'm treating those cases interchangeably. When I talk about 'tables' and 'results' and 'generation', I am referring to the practice of rolling on a random table, reading the result, and incorporating it somehow into the fiction, be it ad-hoc during play or by recording it in a key or adventure plan.

Impartiality & Cognitive Load

When I was preparing the alpha version of Antibor for my recent campaign, I wanted to see how many of the stocking tables in Worlds Without Number I could use. I made six cultures, and then for each I rolled a physical appearance, a set of cultural traits, a religion complete with rituals and temple architecture styles, historical events, and so on. I rolled a ton of tags for cities & ruins the campaign would never visit, secrets for NPCs the players would never spend more than a minute talking to. As you can probably imagine, it was a mess of unrelated data points that were impossible to keep track of. I ended up ignoring most of them & felt bad about it.

A key advantage of a random table (over making shit up) is a sense of impartiality. GMing can never be truly impartial; there is no objective way to translate random details into a playable scenario without inserting some of your own biases. But I've found that building on impartial results in a biased way feels better than negating impartial results in a biased way. Probably because to ignore a roll (even in prep) feels like a slippery slope to fudging them at the table.

With that in mind, I've found that the difficulty (or 'cognitive load') of integrating a table result into the fiction follows this very scientific graph:

Without a randomly-generated prompt, one is left to groundlessly invent material on a whim. (Some people thrive in this mode, but I don't). With a couple of results as a starting point, it becomes much easier to get writing. But push it too far and the complexity becomes unwieldly. With n rolls, there are n + (n-1) + n(-2) ... + 1 potential conflicts between individual results. The odds of incongruity go up and the amount of information one has to track increases accordingly.¹

To avoid situations where I'm tempted to ignore random results, I've learned to start small & avoid generating what I don't need to,² and I try to adopt a consistent procedure for generating any given type of feature. I also try to choose tables that produce 'atomic' details which can be linked up more freely without clashing, which brings me to...

Abduction vs. Deduction

After rolling on a table, the results must be integrated into a shared understanding of the imagined game-world. Since no amount of prep can reasonably describe every detail of a fictional entity, gameplay requires some kind of inference to derive details on the fly. Depending on what information one is starting with, different kinds of inference may be used.

For example, a table used to 'dress' dungeon rooms could have entries pertaining to general categories of room like 'library', 'kitchen', 'torture chamber', or it could list specific furnishings, like 'bookshelf', 'oven', or 'iron maiden'. In the former case, the GM might have to reason deductively about the room if players ask about specific items: 'it's a kitchen, so there's probably a prep surface with knives, and racks of herbs, an oven...'. In the latter case, the reasoning is more likely to be inductive or abductive, starting from a few specific details and progressing to either more details (induction) or a general room category (abduction). 'There's an iron maiden, and a bookshelf -- perhaps this is where apprentice torturers were trained. I'll say there's also a lectern here where the master torturer discoursed about excruciation techniques...'

I'm finding it a lot easier to use induction & abduction. Partly, this is because concrete details are what I as GM actually convey to players, regardless of whether I have a deeper understanding of their causes & context. Having to proceed deductively from a general category tends to be a stumbling block as I inevitably pause to grope around for fitting details to narrate.

I also just really enjoy abductive reasoning; working deductively from a general description has always felt more like busywork to me, whereas with abduction I get to play detective & make sense of the peculiar juxtapositions that the random generators spit out. Proceeding abductively from details also introduces the tantalizing possibility of being 'wrong', if it's possible to be wrong about one's own campaign world. It evokes a sense of ruinous mystery that (I find) tends to get lost when positing an underlying structure or 'truth' and adding specificity.

Luke Gearing's Carcosa review has been on my mind while thinking through this:

Question-raising is both a technique and a stance. Even if the “answer” (according to the author) is contained within the work, by creating distance between them we invite the reader to develop their own answers. When (if) they discover our answer, they are then able to disagree, to synthesise the two, or reject both and find another, novel answer to the questions raised in the text. Questions beget questions. Examples of this can be seen in using generators to ask questions (e.g. Traveller planet generation) rather than provide answers (e.g. SWN planet generation). There is an intersection here too - encounter tables answer the question, “what is here” but also raise the question “why?”.

This was written about McKinney's Carcosa, but also applies to classic modules like B2; the capacity to synthesize unexplained details makes reading & prepping a module engaging and gives GMs a chance to put their own stamp on it.

However, I'm sure this is a subjective preference. Storygames are popular & seem to rely heavily on deductive reasoning -- mechanics will indicate broad abstractions like 'setback' or 'degree of success' and require players and GM to collaboratively embellish them to the desired level of detail. I can see how that would constitute its own kind of fun even though it doesn't do as much for me.

One of the other reasons I didn't vibe with the WWN tables as well as I had hoped is that most of them operate at a very non-specific level (requiring deduction to produce details to use in play), like this table of historical high points I flipped to randomly:

I'd rather arrive at these sorts of broad characterizations by piecing together cryptic details, or never arrive at such a broad statement but still have a list of cryptic details to reference. My ideal 'historical high water mark' generator would spit out things like 'The Solarian Reformation' or 'The War of the 12th Mithril Scepter' or 'The Birth of the Dog-headed Maiden'. Unfortunately, that type of generator doesn't vibe well with WWN's setting-neutral stance, so I get why it's written the way it is. (But it's why I'm ever more committed to writing my own shit).

#game-design #musings #random-tables