The Horus Heresy: A Social Network

Social networks – the ties between people – are a really vibrant area of research. They’ve been used to do everything from examining the collapse of Enron to identifying hidden influencers in social groups. They’ve also been used in fiction to look at how a story is structured, for example tracing which characters appear in the same scenes in Les Miserables.

So what does a social network of the Horus Heresy look like?

I thought this would be interesting for a few reasons:

  • The opening of the Heresy focuses on a lot of characters that while theoretically less important are never the less tightly integrated into the groups they’re in. Being able to visualize that would be interesting.
  • The Legions have a structure to them, rather than that structure purely being something that arises from the narrative. And they’re big enough not to assume that even an entire legion knows one another.
  • You can see how disparate parts of the heresy connect, and how close certain characters are who might never have met each other (for example, Loken and Garro share several connections during Istvaan III, despite never meeting (IIRC).

We’re going to start off with a single book: Horus Rising

‘I’ve never spoken to you,’ Ekaddon said. ‘Never, informally. We don’t meet or mix.

Some Terminology

Networks have some terms used around them, which I’m going to use as well. As such, it’s probably useful to define some of them:

  • Node: The things in the network that are being connected. For the internet, these are computers. For social networks, these are people. For this post, they’re characters in the book.
  • Edge: A connection between two nodes.
  • Centrality: A measure of the position nodes have in the network as a whole.

How I Went About This

Basically, I read the book. Not for content, but for signs of two characters interacting. When they did, I create a node for each character, recording their associated faction (i.e. The Sons of Horus) and their “type”, a brief description of what kind of character they are (i.e. Marine). The basic types (so far) are Marine, Primarch, and Civilian, which is functionally anyone non-Astartes. The factions are any of the Legions that appear, along with “Imperial” for other Imperial citizens, “Xeno” for non-Imperials, and “Chaos” for…well…Chaos. Or more specifically daemons. As they’re interacting, they also get an edge added between them.

Now here’s what’s tricky: defining interaction. For this, I assumed several things:

  • If one person is addressing an audience (for example, Horus addressing the Mournival) the speaker is addressing all of them.
  • References to everyone hanging out (for example, a passage discussing how Saul Tarvitz was quickly befriended by the Mournival) creates a link between every member of that group.
  • Merely being present, but not speaking (for example, acting as an honor guard) does not constitute an interaction. Similarly, members of a squad aren’t assumed to all know each other without interacting (this becomes important later in the heresy with lots of ad hoc squads).

Lots of networks give their edges what’s known as a “weight”, which is a measure of importance. For example, an edge with a weight of 2 is twice as important as an edge with a weight of 1. These weights are often how frequently two characters appear, by scene, chapter, etc. I’m not doing that. I thought about it, but I found this was really ambiguous as to what counted as a scene etc. What is less ambiguous, and will get at largely the same idea, is to treat each book in the Horus Heresy as a scene, and weight by number of books where an edge appears. So, for example, right now Loken and Abaddon are equally connected to Horus, but as time goes on, the connection between Abaddon and Horus will be given more weight. But that will come up in the future.

The Network

horus_rising_network

It’s a complex figure, so lets break it down some. Nodes are colored by their factions, generally in a color corresponding to that Legion – teal for the Sons of Horus, pink for the Emperor’s Children, yellow for the Imperial Fists, red for the Blood Angels, dark red diamonds for the Word Bearers. Imperials are grey dots, Chaos is the black star (in this case, just Samus), and white diamonds for the Xenos (in this case, members of the interex).

Lets highlight a couple things in particular. The first is that there’s clearly some different sub-groups. There’s a dense network of Sons of Horus marines, made up of the Mournival and several other captains. There’s a separate group of Emperor’s Children that largely connects to itself, but is tied to the Sons of Horus via Tarvitz and to a lesser extent Lucius and Eidelon. The very few Blood Angels who are present are tightly integrated, while Dorn and Sigismund are much less so – indeed, while it’s hinted at, neither Imperial Fist actually speaks to Horus, or spends time with the Mournival. Contrast that to Erebus, who is right in the thick of things. There’s also a few distrinct civilian groups. Those toward the top are rememberancers, while the bottom groups are primarily the Imperial military.

This is Loken:

Loken

As you can see, he’s densely connected to basically everyone, which makes sense for this story. He’s one of the bridges between the Legions and civilian groups, and also exceptional because of the amount of marines connected to him – both he and Saul Tarvitz have squads they interact with, while much less is shown of say, Aximand’s squads. Surprisingly, the other major bridge between the civilians and the Legion, that node to the bottom right next to Horus…is Maloghurst. Which makes sense for him being Horus’s equerry.

Breaking Down the Network

There’s lots of different ways to look at a network. One of the simple ways is to look at the degree distribution – that is, the distribution of how many edges go into each node. That’s what’s shown below.

horus_rising_degree

First off, we see that this is a pretty long-tailed distribution, which is common for social networks. The vast majority of characters have four or less connections, and one character has an extremely high number of connections (that would be Loken). Following him are Horus, Maloghurst and Torgaddon. I do think it’s interesting that Maloghurst keeps appearing – he’s rarely an interesting or prominent character in the story. Rather, he’s just always there. Fitting for his character.

Another way to look at this is what are called k-components. These are groups where you would have to removed at least k nodes to break it into a smaller group. Up at the top, at k = 10, we have this grouping:

‘Goshen’, ‘Moy’, ‘Torgaddon’, ‘Horus’, ‘Iacton’, ‘Aximand’, ‘Targost’, ‘Abaddon’, ‘Marr’, ‘Loken’, ‘Sedirae’

So Horus, the Mournival, and several other captains in the Sons of Horus. If you eroded away the rest of the network, this would be its core.

There’s also some centrality measures we can look at: Betweenness and Closeness.

Betweenness is a measure of how central a node is to the network as a whole. It’s the number of shortest paths between all two pairs of nodes that go through that node. Again, the node with the highest betweenness? Loken. Followed by Maloghurst, Tarvitz, and Horus. Those are the “friend-of-a-friend” type of nodes in this network.

Closeness is another measure. This one is the reciprocal of how many steps you’d have to take through the network to get from a given node to every other node in the network. So for example, if Abaddon wanted to give Euphrati Keeper a note, now many people would it have to pass through? If he has high closeness, the answer is not very many. If he has low closeness, the note would have to travel through lots of hands before it reached her.

Guess whose got really high closeness? If you guessed Loken, you’ve been paying attention. This is followed by Torgaddon, Aximand, Maloghurst, Abaddon and Horus. Here, the Mournival is cheating. Loken has extremely high closeness because of his high degree – he’s connected to a huge number of characters through direct interaction, which isn’t surprising, as he’s essentially the main character. But the Mournival is only one step away from Loken – basically, they borrow from all his connections.

One other thing that I think is interesting? How isolated the Remembrancer are. There are 70 characters total, and if you sort by Degree, the first Remembrancer is Keeler at 8 connections, the same as Sindermann. That puts them in 18th place at best. Betweenness is better – Keeler has the 9th highest betweenness, as there are a number of characters who are connected to the network through her, mostly fellow Remembrancers. But it crashes back down again for Closeness, where Keeler returns to 18th place – she might connect a lot of people together, but she doesn’t have direct interactions with a large number of them. Supposedly important characters like Mersadie are even worse – she’s in 26th place for Closeness (below Samus, a bloody daemon), 21st in Betweenness (behind some random characters you likely don’t remember, and barely above the currently disgraced Eidolon), and 28th in Degree, below both Samus and Eidolon.

There’s a feeling in the story that the Remembrancer order is marginalized and ignored, and that’s reflected numerically.

Where Do We Go From Here?

What I plan to do next is to do this for other books, and start slowly digesting the entire Horus Heresy series into a collection of networks, plus a single grand network. This is a side project though, and while getting the data isn’t hard, it is tedious. Some books will be easier – Damnation of Pythos has very few characters, and some will be harder (I’m dreading Unremembered Empire). But I’ll get to it in time, and we’ll be able to see how the story evolves.

Enjoy what you read? Enjoyed that it was ad free? Both of those things are courtesy of our generous Patreon supporters. If you’d like more quantitatively driven thoughts on 40K and miniatures wargaming, and a hand in deciding what we cover, please consider joining them.

 

1 Comment


  1. It’s a very fascinating look, however it does need to be contextualized as literary analysis in the context that all speakers are officially unreliable narrators. This study may be more about the writers than the characters. It did remind me of the fascinating work that goes into metadata analysis though, such as what the NSA does. MIT labs put out a tool so people could run their own email accounts through a metadata analysis and see emergent social networks – https://immersion.media.mit.edu/

    Reply

Leave a Reply

Your email address will not be published. Required fields are marked *