Big decisions you may not be consciously making.
All World-Building is Modeling
First, I want to propose that all world-building is model building.
When I say “model” you may think of something technical, like a climate model running on some super computer. But there are many kinds of model. They can indeed be mathematical, but can also be purely mental, drawn on paper, or made of clay. A model is really just a (theoretical) understanding of a system.
Many people turn to computers when making worlds, because computers can track details and apply rules in uniform ways. And indeed, a lot of what I will present here is borrowed from scientific modeling — but mostly because these are concepts that I think can help any world-builder, regardless of the tools being used. Computers aren’t required.
Models can be Spatial
Often cartography — making maps — is at the heart of a world-builder’s enterprise. This means the model is spatially explicit: items of interest all have a position, and possible positions are continuous in space.
The alternative is a spatially implicit model, where physical relationships might be known but are not mapped out. For instance, “the capital is on this continent, in this province.” This is often used when world-building is in service to another project, like a novel (although a world map has become de rigour in fantasy, thanks to JRRT).
For the rest of this piece, I’ll more or less assume a map-based, spatially explicit world-building exercise, but both approaches are viable.
Detail versus Scale
People often talk of detail in worlds (and games and novels). I think it’s more useful to think in terms of scale: what scale of features and events are you describing? What scale are you modeling? This depends on what matters to you, and what the purpose of your world will be.
You can model at multiple scales, and probably will; but you should still match the scale to the item of interest. Geologic process occurs on a very large scale, and you would be wise to model it that way — to think about continents, and not square kilometers of soil. Settlement patterns and the street layouts of cities, meanwhile, will be handled more finely.
By thinking carefully about scales, you can help handle the different natural processes that create a world, like volcanic eruption and technologcial diffusion. Smaller scales are “embedded” in larger ones, including chronologically: many small events can occur within the span of large ones. This can simplify your job, because the big events create the context of the smaller events.
Differences in time scale can also create a process for you as a world-builder. You can tackle continents first, then mountains, then rivers, then settlements, etcetera — with each stage being dependent on the ones before.
So the question is not about whether to have details, but what kind of details, for which processes, and with what interactions?
Process or Just Results?
This final item is about the internal dynamics that happen “behind the scenes” in your model, separate from output. You might like to have some continents, for instance. One approach is to learn about tectonic plates — drift, uplift, and so on — and then model those dynamics in your simulation, hopefully with realistic seeming results. The other main approach is to use some procedures that make realistic continents even though they have nothing to do with realistic geology. You might, say, splatter paint on to the floor, or program a fractal.
This distinction can be talked about in many ways. Some will say “top down” versus “bottom up,” or scientists may talk about “phenomenological models” versus theoretical ones. I’ll just stick to process- and results-focused.
A lot of technically-minded people assume that the process-oriented approach is always best; that the more realism you can pack into the model, the better it will be. The problem is that simulating real-world processes is extremely complicated. It takes a good deal of domain knowledge, hard work, and (computation) time. And it will probably require a lot of tweaking, parameter-setting, and futzing to actually produce results that you like — if you ever can. Modeling is hard, and scientists don’t even have complete theories for most processes you’ll be trying to recreate. It could be a lot of effort for little reward.
Results-oriented modeling has a big upside: you’re focused on the level of realism you care about. If you want nice coastlines, you can make sure that happens, because you’re flexible on methods, and have no prior commitments. It can also be much faster, and requires less background research.
That said, all models are really a hybrid: nothing is purely process-oriented, or purely results-oriented. You won’t ever be modeling electrons on your planet: it’s simply the wrong scale for what you care about. So we’re talking about scale again. The key question is whether you take the features you’re concerned with and go one level deeper to create them, or just stay at the surface-level, and find a process that works. You can make different decisions for different features, of course.
After talking down process-oriented modeling above, let me say a few words in its favor. The best thing it offers, in my opinion, is serendipity: when you create processes that work in a more hidden way, you can get more results that you never would have expected. And surprising features don’t have to be silly or unrealistic this way: if the process is basically sound scientifically (even if very abstracted), the results will be too. Process-oriented approaches can also be a little more robust: you can apply them in different scenarios, with different starting assumptions (a world of all islands! a world with one huge continent at the north pole!) and they’ll still work. Results-based processes often end up being more finely tuned to specific situations; change the situation, and you’ll have to find some new methods.
Hopefully, this will help you make some thoughtful decisions about the models (worlds) you want to create.