This is an overview of the rules I’m using for my ‘incremental worldbuilding.’ Basically, starting from a blank slate, I make small changes over time to hopefully add a sense of weight to the worlds I create. This is specifically based on a pre-writing, pre-agriculture, pre-history society - when I get to an appropriate point, I’ll create rules to implement early civilizations forming and agriculture, but I’m nowhere near that yet.
I use ChatGPT extensively in this, as it both simplifies the idea creation process and occasionally even provides interesting things that I hadn’t considered. It also is a way to simulate language change in a way that is at least plausible, if not realistic. One caveat to that, though, is that I’m using GPT-4 for this. GPT-3.5 is significantly less capable, and, particularly for the language process, which is the only part where I actually rely on ChatGPT rather than just use it to help my own idea creation, just doesn’t meet the mark.
I’m making rules (and rolls) for my worldbuilding for two reasons. First, it adds randomness, which I hope will make the world feel realer and more grounded than if I decide everything each time. As I go through the process, a roll might pick an option that isn’t the one I wished for, or thought was coolest. The way unexpected results combine together will create something that feels more real than if I hand-crafted everything.1
The second reason is for fun. I like it when I get surprised by something, when I have a bunch of options to roll between and I roll the one I thought was least interesting, but then I turn it into something more interesting in a cool way.
Below, I will describe how I go about a turn of worldbuilding. This is not a complete description, I don’t go into detail about each step I take in this (although at some point, I may add in additional details in that way). These rules have changed, and will continue to change as I go through the process and find additional interesting things to do and alter. It will not necessarily be the exact same process I used for any particular turn of worldbuilding. At the very end of the post, I have a few comments for anyone who is interested in using these rules to try it out themselves.
Turn Order
Random Event: Roll on the random event table, then roll to see what region it occurs in
Prominent Individuals:
In any region that had a prominent individual last turn, determine how that individual is remembered and/or how the remembrance has changed, then move them into Myths/History.
Roll a region. If that region has anything in myths/history, see how that remembrance has changed over time.
Roll a region. That region has a new individual come to prominence. Conduct a basic language development on that region.
Expansion: Roll 1d2. That many regions will expand to a neighboring region. For each, roll a region, then roll among possible destinations. If the region is populated, that region will migrate to a nearby region, and the activated region will take their place (waves of migration pushing previous waves away).
Biome Generation:
Generate a biome for the regions that have been expanded into. Every nearby biome will contribute one feature to the new biome, and there will be two taken at random from different already existing biomes as well.
Develop new words to describe all new creatures.
New regions and source region each undergo basic language development
Language Development: Roll a region. That region undergoes language change in all categories
Advancement / Development:
Roll a region to determine which undergoes technological/social/cultural advancement.
Roll a 1d3 - 1 is technological, 2 is social, 3 is cultural. The selected region will have a minor advance in that category.
That region will also undergo basic language development.
Random Event List
An animal migrates from one region to another
A species dies out
A new species becomes prominent
A new variant of a species becomes prominent
A new plant becomes prominent
A new resource is found (possibly related to technology - technology advances as well?)
Cultural schism: A region splits into two distinct cultures (possibly ignore/re-roll if it’s a biome that can’t support a large population)
A loan word is transmitted from one region to another
An additional round of linguistic change occurs - full language development in the rolled region to all categories
Society devolves - some complexity of society is reduced
Society evolves - complexity is added
Society changes - some aspect of society changes
Culture devolves - some complexity of culture is reduced
Culture Evolves - complexity is added
Culture changes - some aspect of culture changes
Technology devolves
Technology evolves
Technology changes - a new or different take on the same thing (not sure if this makes sense - maybe a change in priority - hard to distinguish from devolving or evolving)
Language evolves complexity - add particles, tenses, etc.
Language devolves - remove particles, tenses, verbs
Culture massively changes - massive rapid changes across the entire culture
Major linguistic shift - run language development twice
Major societal shift- massive rapid changes across the entire societal structure
Genetic change - develop new feather colors, for example
Individual figure rises to prominence (shapes culture, later turned into myth)
Major natural formation/landmark adopted into cultural beliefs
Mystical/unexplained phenomenon causes new beliefs/myths
A resource is depleted
Local Natural Disaster (flood, drought, etc.)
Local legend about place, event or individual becomes widely shared
Major Event of Planet-wide importance - instead of rolling to pick a region, roll to see which of these events occurs: 1. Another species gains sentience. 2. Global cooling. 3. Global Warming. 4. Major volcanic eruption. 5. Meteor impact. 6. Extra-terrestrial visitation (probably without contact, probably leave artifacts somewhere). 7. Major plant/animal die-off. 8. Major disease. 9. Sea-levels rise. 10. Sea-levels fall.
Planet-wide advancement - every region goes through some change: Roll: 1. Language; 2. Culture; 3. Society; 4. Technology; 5. Genetic
A region is subsumed into another nearby region (only applies to regions from the same biome)
Biome Generation
Overall, for biome generation, I want to create the following things for each new biome:
1 Major Predator
2 Minor Predators
1 Scavenger
1 Large Prey animal
2 Medium prey animals
2 Small prey animals
2 Staple foods (for the animals and/or the Eekraw)
1 or 2 local resources
For the very first biome, I just entered the following prompt into ChatGPT. I regenerated the prompt four or five times, picking the best or most interesting options.
You are a brilliant and creative science fiction author. I'm worldbuilding a sc-fi world. I'm going to create creatures for each biome in the world I've made. Help me brainstorm the following things for a tropical rainforest biome. Be creative and think through how such an ecosystem would function and what creatures would live there.
1 major predator, 2 minor predators, 1 scavenger, 1 major prey animal, 2 medium prey animals, 2 small prey animals, 2 staple foods, and 2 usable resource
Be creative, varied and avoid cliches. Avoid direct analogies to terran creatures
As the turns continued on, I’ve made changes to make the biome generation process less time-consuming. It was cool the first time, but much less interesting the tenth time, and it’s not the part of this that I have the most fun doing.
Now, if it is the same type of biome as a nearby existing biome (i.e., they’re both temperate forests), I take that nearby biome and then alter it by the following procedure:
Roll for a feature to be replaced by one from each of the neighboring biomes.
Roll for one or two features to be replaced by features from a random existing biome
Roll for one or two features to be replaced by a variation of itself
If there are no similar biomes nearby, I create the new biome by the following procedure:
Roll for a feature to be added from each of the neighboring biomes
Roll for one or two features to be added from a random existing biome
Fill the remaining slots with newly created features.
I’ve started using random tables (such as the ones from Stars Without Number by Kevin Crawford, or Star Trek Captain’s Log from Modiphius) to generate features about the creatures, then put them into ChatGPT. I think this will prove an effective change that makes things much easier, but I haven’t fully implemented it yet.
The last part of biome generation is determining what the locals call the new creatures they encounter. I figure that out by the following process:
See if any creatures are similar enough to ones they already know to be called by the same name. E.g., the many variations of the Varkee (avian predators) or the bioluminescent beetles that exist in many biomes.
See if any creatures are similar enough to ones they already know to be called by the same name with a simple modifier. E.g., a tribe that used to live in a biome with Varkee, now sees smaller avian predators and calls them Varkee’tiy (small Varkee).
Finally, assemble a name from combining various words from the lexicon together to describe the creature. This does not need to be a short or even reasonably length word - part of the process is for language change to shorten words over time. E.g., the Kra’tir’skikka from Region 7 during Turn 4 (the word means defense-harmless creature) has turned into Kraskia by Turn 6.
Language Development
There are two types of language development - basic and full.
I do basic language development for every region that gets activated in any of the turns. For example, during the Prominent Individual section, the region that is rolled will also do basic language development, and during the advancement / development section, that region will do the same. Basic language development consists of rolling a random category of language, and changing all the words in that category.
I do full language development only during the language development part of the turn order. For full language development, I change every word in the lexicon for that region.
For both types of language development, I put the following prompt into ChatGPT:
You are a language history simulator. You simulate the changes of language over time based on the Principles of Linguistic Change by William Labov. You are simulating the changes of a sentient avian species as described below. This tribe lives in a __ biome. Given the lexicon below, simulate the changes to the language over time.
I then put in all the details of the Eekraw, the societal overview of this region, and the language, and end by typing “Simulate language development for this tribe of Eekraw.” I usually get three to five usable word changes from this, across the entire lexicon. I’ll do this two or three times, resulting in somewhere around five to ten words changing over the entire lexicon.
Then, for basic language development, I’ll roll for a category of words from the lexicon. For example, I might roll (2) - cultural words. Then I’ll use the same setup as above, but instead end with “Simulate language development for every word in Category 2.” This will almost always result in ChatGPT making changes to every word in that category - however, sometimes I have to regenerate to get a useable output.
For full language development, I’ll do the same thing, just for each category in turn, meaning that every word in the entire lexicon changes.
Sometimes I’ll also run it with the following end prompt: “What words might undergo semantic broadening/narrowing, amelioration, perjoration or metaphorical substitution?” to add some extra interesting changes to the process (though ChatGPT often does some of those on its own in the above steps without being specifically prompted).
I don’t always take exactly what ChatGPT provides. Sometimes, I’ll take the concept that it provides but change it, or do the change it suggests but to a different word, or to only one of a set of words with similar parts.
At the end of this, I have a language that has materially changed from its predecessor, both in specifics of words and speech, and often in the meanings and connotations of words as well.
There are obviously serious limitations to this process in terms of how well it represents real language change - not least that all linguistics research is about how humans pronounce words, not how an avian species would. Nonetheless, from the perspective of creating an interesting language for the purposes of worldbuilding, it creates some awesome results.
Using These Rules
I have a few thoughts for anyone who is interested in using these rules to do their own gamified incremental worldbuilding. First of all, that would be super cool, please do and tell me how it goes! Secondly, this not a time-efficient or effort-efficient way to do worldbuilding. Use this for fun because you’re enjoying the process, not merely because you think your worldbuilding would benefit from it - I think this would be excruciating if you weren’t already excited by and enjoying the basic process. I think of this as basically a Solo RPG that happens to be centered on a planet, not on a character.
If any reader actually decides to try this out, I have a few recommendations. The only things that need to be generated beforehand are a map, and a bare amount of detail about the creatures you’re following (it’d be much easier if they’re human - I partly wish I hadn’t gotten too ambitious and made an avian species). I used a cool software program to create a really detailed map that includes biomes generated directly by the program, but you don’t have to do that - you just need a basic map with a general idea about different areas in the map and the biomes of those areas. You can add more details, both about the planet and about the sentient creatures whose development you’re following (assuming non-human), as the question becomes relevant.
For the first two or three turns, I recommend you skip the Prominent Individuals and Advancement / Development rounds and either skip or greatly limit the options for the Random Event Roll. This will allow the different tribes to spread out a bit and differentiate before adding these changes, making them more different from each other. On the other hand, you may prefer not to do that, so you can see how the story of a character from the first region changes as the tribes spread out.
One cool thing to try would be to have two separate sentient races, each expanding towards each other independently on the world. The question would be how the interactions can be gamed out when the two species meet finally.
At some point I may write this up as an actual rulebook, but for now, all you get is a description of how I do it.
This, in fact, is the biggest difficulty if you try to use ChatGPT to work as a Game Master for a roleplaying game or anything like that. The way the generative AI models work, it will always pick ideas that are more obvious. The first time I used it, that wasn’t apparent and I was just blown away by its capabilities. But as I used it more, I found that to use it most effectively, I need to add some randomness from outside. Otherwise the same situations and actions keep repeating.
For generating specific characters or factions in a RPG setting, one trick that I’ve found useful is to steal a concept from Solo RPGs. Basically, you have a list of words and roll between them. Then, when you ask ChatGPT for a new character, you don’t just say “A human barbarian.” If you do that, it’ll always describe a pretty stock character who wears furs, enjoys fighting and despises soft city folk. If you roll from a table of random words, however, you can say something like, “Describe a human barbarian that has characteristics related to the words ‘zeppelin’ and ‘symphony,’” which will give a very different result than if you use the words ‘glacial’ and ‘mauve.’ The more random descriptive words you add, the more unique, though it can become a little bit too much if you push it too far, and anachronistic words can add anachronistic elements if you don’t work hard to avoid them. I like adding in a positive and negative character trait and a word of physical description along with the two random words to get particularly varied characters.
I haven’t used that particular technique in this worldbuilding project, however. Instead, I’ve been doing a process where I generate a number (10 or more) of ideas, pick the best, and then roll between them to generate randomness. This works a little bit better in this context, where I’m not dealing with individuals but with changes to cultures. However, I have noticed repeats of the same ideas come up in generations with different contexts, so I may start doing a similar random word insert at some point. It’s a constant fight to keep things being plausible, but prevent them from converging on the obvious.
I like this philosophy on worldbuilding! I might have to try some sort of rules/structure like this for my ongoing worldbuilding project.