As storytellers, we’ve been educating our audiences to be able to handle more complex narratives over time, from embedded narratives and examples of metalepsis to explorations of human nature. Another development is the shift in focus from single stories to story worlds.
Nowadays, fresh IPs for not just games but also movies and series are pretty much expected to take place in story worlds, showing multiple narratives from diverse characters – often morally grey and complex – to yield more entertainment and effect from the investment in developing that IP. I would argue that since the early 2000’s, but especially in the past decade during which games like Dungeons and Dragons have become more popular than ever, worldbuilding has become a serious pastime of professionals and amateurs alike. Worlds that belong to the Assassin’s Creed series and Warcraft continuously expand, so that players can consume them endlessly.
This is the first article of a four-part series that shows how networks and sociology (and more specifically, theories on the sociology of art) can be applied as a worldbuilding method that yields meaningful and dynamic worlds. These articles support my GDC2021 talk ‘Connected Worlds: Building Dynamic Storyworlds Using Network Theory‘ by offering more details, substantiation and calculations behind the mechanisms.
Throughout these articles, I am using yEd Graph Editor to build the examples and models that illustrate how my method functions. Please feel free to use whatever (mind-mapping) software you prefer – just bear in mind that worldbuilding also requires content, so use a tool that can link to the content you created.
Back to worldbuilding!
We all love it.
But what I noticed a few years ago, is that the act of worldbuilding is poorly defined. There doesn’t seem to be a proven or preferred way to do it, nor a method that reliably yields outstanding results. Instead, when discussing worldbuilding, most people tend to adopt an ‘anything goes’ attitude. Which is fine – don’t get me wrong – but as a professional I was looking for something more concrete to wield.
So, when I started to think about developing a worldbuilding method, I looked at three ways people tend to worldbuild: by creating narratives, drawing maps, and composing timelines.
Let’s examine what these elements actually are and how they overlap. The images used in the paragraphs below are from the outstanding game Heaven’s Vault by Inkle Studios.
A narrative is a story told through a character’s perspective. They have a beginning and an end. Narratives nearly always follow a plot, which is defined as a series of events. Events are meaningful or consequential moments that affect characters or decisions made by characters and take place at a location: some place in the story world. A character’s narrative intersects with that of others, and sometimes runs parallel with a companion’s narrative for a while.
For instance, it’s not uncommon for mentor characters, who teach young heroes about the ordeals that they are about to face, to have a respectable reputation themselves in the story world, earned by performing mighty deeds in the past. When building a story world, it’s easy to connect the narratives of the mentor and mentee by interlinking their individual narratives.
Maps are systematic representations of geographical elements that show how locations are placed in relation to each other, including how they are connected. They disclose where interesting or major events occurred, such as battles and adventures, but also where people live.
A map is usually synchronic, meaning it shows you the state of the world at a single moment in time. However, maps can become diachronic by showing the passage of time by adding dates and years to events.
Lastly, there’s timelines, or perhaps histories, which show a sequence of events. Linking events can reveal correlation or even causality. As we’ve mentioned under Narratives, sequences of events that meet certain conditions are called story plots.
Shift to networks
Looking at narratives, maps, and timelines, I argue that you can generate more of these by using three basic story elements: characters, locations, and events. Once you organise these elements as nodes in a network structure and draw lines between them, you are able to declare their relative position to each other and include additional information that way. The connections between nodes are sometimes more important than the nodes themselves. Once all nodes have sufficient qualitative connections to others in the same network, a vibrant world begins to take shape.
This could help prevent the so-called theme park worlds that we may find in games. In these types of worlds, some clusters of characters, locations, and events barely seem connected beyond their own little zone, making all of these clusters feel isolated, too systemic, or sterile.
Real worlds, on the other hand, are messy and convoluted. All things noteworthy in our world boil down to:
Or, paraphrased, a Character involved in an Event at a Location. Let’s have a better look at these three types of basic story nodes.
Characters are the active agents in the world. They incite change and harbour values. Their decisions and actions cause events to happen and reshape locations as time progresses. When worldbuilding, it may be tempting to design factions, cultures, and tribes, but I recommend you focus on individuals. Decisions are made by individuals. Players and readers empathise with individuals, not masses. Identify your leaders, heroes, martyrs, and villains in your factions, and ask yourself how their actions shape the future of their people.
You could consider items/objects a subcategory of characters, or maybe they behave similarly to characters. Consider legendary items such as Excalibur, The One Ring, or Mjölnir – these items interact with other characters and have stories focus specifically on them. You could include such items as characters. Other items that have less personality or that never switch hands might be better fitted as an attribute of a particular character, such as a war horn or a favourite rifle.
Increasing group sizes increases the chance of conflict exponentially. If you begin to add more and more characters to your network, families begin to take shape, rivals are introduced, friendships forged, and ambitions clash. Characters create drama as soon as you draw the connections between character nodes. The drama is locked in the description of their relationship, and with it the values and morals that the characters have. More importantly, by drawing connections between characters nodes, you tend to focus on the differences (in background, ambitions, personalities, and so on), making it even clearer how conflict either originates or is justified between characters.
The more characters you insert into your story world network, the more drama (or story) you will generate.
Locations are usually clearly defined, yet meaningless without context. Connect them to other nodes, and they become battlegrounds, places to visit, or destinations to reach. They are the spaces in which characters move and live, and where events happen. They can be man-made or natural spaces. Locations don’t have to be physical spaces; they can include elemental planes, astral dimensions, or virtual cities on a hard drive. I recommend you focus on locations that are meaningful to characters. You can create endless locations and include them in vast maps, but if no events happen there, or if they are not linked to any characters in the world, they may be either obsolete or inconsequential!
Continue to add locations and you’re creating a map. Particularly when you draw connections between locations.
The more locations you add to your network, the more geography (or map) you’ll end up with.
Events can involve multiple characters, or just one. They could even be character-less, such as a volcanic eruption, but should always affect a character, lest it has no impact on the world and, therefore, is of no consequence. Events are not identical to moments, however. They can span years. The melting of the ice caps could be an event in your story world, despite it taking decades if not centuries.
Events happen at a specific moment in time. This means that if you organise your events along a linear timeline, they may become a plot. If events are connected to each other, you are introducing mutual logic to the events, most likely correlation or causation. It’s indeed probable that your world will end up with multiple parallel timelines: series of events that do not directly impact one another, but that do occur simultaneously in your world. The coronation of the orphaned Prince, the slaughter at the Mudgate, the marriage of Olaf and Theresa; events impact characters and may become a basis for their shared relationship, therefore leaving an impact and meaning for your world.
Getting breakfast usually isn’t marked as an event. Events should deserve a newspaper headline, generally speaking, but it doesn’t need to be front page. For those of you craving a more formal approach, have a look at how Natalie Heinich defines events in her sociological approach to networks.
Events can include smaller events. The Siege of Minas Tirith (in Tolkien’s Lord Of The Rings) for example includes several (sub)events:
- The siege ram Grond breaks through the city gate
- The charge of the Rohirrim on the Pelennor Fields
- The death of King Théoden
The more events you add to your network, the more timeline (or lore) you create.
Building the network
Let’s put these nodes in context. You could start with a blank canvas, but I don’t think it’s likely that you ever really start from nothing. You probably have a few characters, stories, locations, or events in mind. So, you just need to express them as nodes. You may end up with something looking like this.
In this hypothetical structure, you can see that all connections are possible between nodes. As a next step, you can add new nodes, or possibly entire new clusters of them. To make sure that they feel embedded in the world, draw a couple of connections to other nodes. When a child is introduced, they won’t be linked to their parents alone, but also to their house, events that happen at home, the friends they make in the village, and so on. Don’t overdo it and keep the connections relevant and meaningful; as a rule of thumb, try to draw four connections per new node you insert. Alert readers may have picked up that this increases the connection count for other nodes as well, so by adding new nodes, you add to the interconnectivity of the entire network.
So, add one more character, location, and event, and you may end up with this.
You can also introduce an entire new cluster (A new village? Another age? Another faction?) and link them up in a comparable manner.
But these are all theoretical examples. Let’s look at a slice of a world you may be familiar with.
Example 1: Riverwood
Riverwood is the first village you encounter in Bethesda’s TES V: Skyrim, right after you’ve finished the tutorial. It’s not a large settlement and only has a few characters offering a handful of quests. I consulted The Elder Scrolls Wiki for Riverwood’s names, layout, and images
Here is what Riverwood looks like as a network model:
I included the download link to the graph file at the bottom of the article, so that you can play around with this representation yourself. I had to cut off any further connections after the first degree beyond the last node that has to do with Riverwood, lest I’d be make a representation of Skyrim’s entire world.
As you can see, there are mainly characters and locations in this network, very few events. This has to do with the village being relatively static unless the Player Character takes on quests or shakes things up in another way, such as attacking the villagers. Other events aren’t defined in the game lore, such as the arrival of key characters to the village, or the (de)construction of prominent buildings.
Upon closer inspection, this network has a couple of tightly interconnected clusters, which are only loosely tethered to other clumps of nodes. These clusters are the households that make up Riverwood. The blue group at the top are the various locations that surround Riverwood, or that connect Riverwood to other places in Skyrim. It’s interesting to see that these locations don’t have any characters or events connected to them. The locations simply exist.
Since the image of the Riverwood network isn’t super clear, here’s a detail from the most connected part of the network structure:
It’s now apparent that both Delphin and Esbert are well-connected characters, whose links extend beyond Riverwood. That shouldn’t come as a surprise, since these two characters are part of the main quest line in Skyrim. Note how the model shows them associated with various locations. This, combined with the corresponding events, shows how they move across the world over time, although I didn’t show how events follow or cause each other in this case.
If you set out to develop the rest of Riverwood, you might want to take any node that’s already present, and insert new nodes, or nodes from other parts of the world, that connect with them. Sigrid could have an aunt living in the city of Riften1. Maybe Hod & Gerdur’s house was built by the parents of a mage from the College of Winterhold? Branching out by connecting nodes to other regions of the world help show that every part of the world exists beyond their immediate surroundings.
1 I just want to note here that this could already be the case, but the connection isn’t experienced in the finished product. And that is fine! Including every connection in a project would make the final result feel overtly messy, chaotic, and impenetrable. Much like the real world, sure, but not fit for a world that people intend to keep track of.
Limitations and opportunities
This first example clearly shows that the network model is no substitute for content. You still need to create the characters, what they do and how their relationships to other nodes are formed – and the same goes for locations and events as well. So, you’ll need to supplement this structure with other documents, a wiki, or insert the content directly into the network, which you can do if you’re using software like Twine, Articy:Draft, or yEd Graph Editor:
The benefit of building your worlds through networks is that it is immediately apparent how nodes affect each other. If the connections in the model seem too messy, you can also select any node to see what its immediate neighbours are. This feature makes it very convenient when making changes to nodes, because you can tell how these changes ripple through to other nodes.
So, in addition to giving you tools to ensure that your worlds are built more cohesively, this approach gives you the tools to track your network and how its nodes connect to one another. It gives you more insight than links on a wiki-page.
I also want to talk briefly about how these structures can benefit from introducing hierarchies to the nodes.
Using arrows to show the direction of a hierarchy, connections between characters can easily create family trees in which you can include generations of offspring. It also allows you to show how the power dynamics between characters are defined (which may of course change over time). Introducing hierarchy to events makes it apparent how events flow from each other, resulting in timelines you can track. Hierarchies in locations may show if you are ‘zooming in’ on the map, or ‘zooming out’. This also prevents you from having to draw too many connections between location nodes. Nodes on the same level may refer back to their parent node, so that you don’t have to draw connections between every building in a street (though it can be helpful to connect buildings that are direct neighbours).
Example 2: The Marvel Cinematic Universe (phase 1)
Connecting new nodes to existing characters, locations, or events is a method Marvel Comics uses when they introduce new superheroes or villains. They feel instantly integrated in the world after they are linked to characters that we’re already familiar with, or other previously known artefacts from the Marvel universe. It creates the illusion that this character belonged in this world all along. Cousins, college roommates, childhood friends, employees: Marvel characters all seem to know each other even before they received their superpowers. And that’s smart! This way, you can take the (supposed) existing histories and their related drama, and create new storylines and interactions from that. You don’t have to set up a meeting or first clash.
The Marvel Cinematic Universe (MCU) isn’t any different. The post-credit stingers are most often used to introduce new characters or events by linking them to familiar faces. General “Thunderbolt” Ross meeting with Tony Stark in a bar at the end of The Incredible Hulk. Agent Coulson finding Mjölnir in New Mexico at the end of Iron Man 2. And Loki watching Dr. Selvig talk to Nick Fury about the Tesseract at the end of Thor. They all link new and existing characters and events to one another, while setting up new stories at the same time, thus expanding their world.
I also recreated the MCU’s phase 1 (all movies until the first Avengers movie) as a network story world:
This chart shows all connections between the nodes and is admittedly awful to read. Fortunately, yEd also allows you to clean up the model in whatever way you like. I’m condensing several connections into single lines, so that the nodes themselves become more legible, but I recommend that you download the yEd graph file if you want to have a better look. The download link is at the bottom of this article.
Every new character, location, or event (that has a noticeable impact on the world) that is introduced in the MCU’s later films are integrated by connecting them to more than one other existing node. I think this is part of the reason why the MCU feels so coherent and logical.
Network structure and shapes
When you’re drawing your connections between nodes, it might be tempting to interconnect all nodes as much as you can, like this.
It may sound counter-intuitive, but this doesn’t result in worlds that feel coherent. The household clusters I mentioned in my discussion of Riverwood? You actually want those; tightly knit clusters in which individual nodes form bridges to other clusters. That’s how you keep your network manageable, but it’s also closer to how social networks in the real world work. So, you want your structure to be more like this:
This is a much more organic way to form communities and node clusters. Every node is connected to all other nodes anyway, using neighbouring nodes as bridges.
If you’re wondering about how to balance the degree to which nodes should interconnect, it’s helpful to consider the idea of the ‘six degree of separation‘, or how everyone is only six symbolic handshakes away from any other person in the world. Does the same apply to your world? Are all nodes no more than six steps removed from each other? If they are removed too much, think about drawing a few bridge connectors where it feels logical. If they are too closely connected, consider cutting some ties to create more clusters like in the example shown above.
Example 3: The Return Of The Obra Dinn
My final example covers Lucas Pope’s excellent The Return Of The Obra Dinn. In this game, you are exploring a deserted ship as you’re trying to figure out what happened to its crew and passengers. You visit the exact moment someone died on the ship by using a special pocket watch on the location of their deaths. These scenes, frozen in time, should give you sufficient clues to determine the identity of every individual on the ship as well as their fate.
Characters? Events? Locations? A perfect candidate for a network model!
Again, I’ve condensed the connections here in single lines as much as I could. And here too, you can download the yEd Graph file at the bottom of this article.
If you haven’t played the game yet, beware of small spoilers ahead!
What’s interesting about this model is that all events are linked with direction. They form a single timeline that you can follow, so every event node shows you what precedes and follows it, but also what characters are involved in the event.
The locations are organised in a very particular way in the model. This was intentional, since the blue squares are laid out in an analogous way to how the in-game map works. This illustrates that you can commence your mapmaking while building your network, moving around nodes when needed, but keeping an eye on how they are connected to the other nodes.
Whereas the networks for Riverwood and the MCU took the characters as anchor nodes and organised the other nodes around them (for your own sanity, please use software that allows you to reorganise your structures automatically), I elected to manually shape the structure for the Return Of The Obra Dinn.
I first created all the nodes and connected them according to the events that transpire in the game. Then, as stated before, I used the existing maps to structure the locations of the game. Next, I placed the events – shaped as a continuous timeline – around the reconstructed map, so you can ‘read’ the event progression like a book: from left to right, top to bottom. Since every event is linked to a death (or event), I placed the affected character directly below their corresponding events. Finally, I recoloured some connector lines to include additional information (marking those killed in an event in blue, killers in red, and those that escaped in green).
You can still see how much more clarity you get from manually organising your network this way, although I admit that it will be an ordeal as your world continues to grow. It’s also a valid strategy to manually structure only certain sections of the world, where you can benefit from that process the most.
I hope this article shows how you can benefit from creating worlds as networks. It’s been several years since I first started using it for my projects (both freelance and personal) and it made a noticeable difference to how coherent and congruent they feel. It also made it remarkable easy to track changes and their ripple effects to other nodes.
If you create a network you can collaborate on as a team, you’ll find that you’re expanding worlds at an incredible pace without ever losing oversight of which sections are underused or poorly developed, and which sections are possibly too strongly represented. In other words, it helps you build organic and dynamic worlds.
I’m really looking forward to hearing what you think of this approach. Does it make sense to you, or does it sound needlessly convoluted? Feel free to reach out to me over email or Twitter or leave a message below. I’m always looking to refine the method using feedback from my peers.
You can download all yEd world network files used in the examples by following this link to a Google Drive folder.
More articles in this series
Part two of this series: ‘Connected Worlds: Measuring world connectivity using graph theory’ is unfortunately slightly delayed due to me falling ill this week, but should be up in a couple of days. That article will cover how you can use graph theory (and math!) to measure which nodes are most strongly connected, which ones are key components in your network, and what areas of your world might need more love and attention.
Part three (‘Connected Worlds: Introducing and managing relative power’) and part four (‘Connected Worlds: Values and orders’) should go live soon after that.
Check back in a few days for links to these articles.