Rohan Kapitány, PhD

Fooling around with Graphs (Wiki edition)

I’ve been fooling around with graphs, lately.

There’s a few reasons why.

First, point-estimates of a given construct are functionally incapable of representing the complexity of the individual. But even so, we don’t usually use point-estimates from the individual, but estimates derived from population means. This is known as the ‘ergodic fallacy’ (see: Speelman, Parker, Rapley, & McGann, 2024). While graphs do not strictly resolve this issue, they can provide a far more complex way to represent an individual, as well as representing the complexity of a sample. But that’s neither here nor there for the current moment.

Second, graphs seem likely able to provide interesting insights into a given phenomena from a cultural evolutionary point of view. I mean, phylogenetic analysis are a kind of graph, and we all accept that those - in principle - have value.

Third, as an unfunded scholar, I’m always looking for ways to be able to do research for free, and a lot of data that I’m able to access for free has network properties baked in. For example, wikipedia.

GRAPHING WIKIPEDIA

You all know what Wikipedia is. I won’t go into detail here, suffice to say that Wikipedia is a good thing. A crowning accomplishment of humanity, in my humble opinion.

So I built a shinyapp (using Claude) that scraped Wiki. This is actually the second scraper I made, and it’s the simpler of the two. This scraper takes a single seed wiki page, scrapes all the outlinks on that page (excluding things like ‘talk’ and ‘edits’), and then scrapes all of those pages, too. (It can go deeper, but it tends to be fairly unintelligible at that point, because most things are within three steps of something like ‘The United States of America’, and any inferential potentially goes away). Functionally, it finds all pages/nodes that are within two edges of the seed. In a sample graph , the seed will be the centre.

For example, here’s a graph for ‘Dungeons and Dragons’ (with 200 of 72000 nodes visible).

D&D is clearly in the centre, and there are two main communities/clusters. On the left we have things that are strongly related to D&D, and on the right we have things related to literature and genre. Of particular interest in J. R. R. Tolkein, who serves as a bridge between the two.

If we look at betweenness centrality, which is measure of how often a node acts as a bridge between any two points, ‘Dungeons & Dragons’ is king. The top nodes on this metric are:

1. Dungeons & Dragons
2. Fantasy
3. Halfling
4. Orc
5. Magic: The Gathering
6. List of Advanced Dungeons & Dragons 2nd edition monsters
7. Internet Speculative Fiction Database
8. D&D Beyond
9. Chimera (mythology)
10. Demon
11. Gary Gygax (Had to include Gary here, even tho he didn’t make the top 10).

The closeness centrality, a measure that computes how easily a single node can reach any other node in the graph, looks very similar:

1. Dungeons & Dragons
2. Halfling
3. Fantasy
4. Orc
5. List of Advanced Dungeons & Dragons 2nd edition monsters
6. Magic: The Gathering
7. Gary Gygax
8. D&D Beyond
9. Monsters in Dungeons & Dragons
10. List of Dungeons & Dragons 3rd edition monsters
11. Margaret Weis (the creator of Dragonlance)

And the eigenveector centrality, which is a measure of a node’s influence as a function of the importance of it’s neighbors (i.e., nodes are more influential when connected to other influential nodes), is as follows:

1. Dungeons & Dragons
2. Magic: The Gathering
3. Forgotten Realms
4. D&D Beyond
5. Gary Gygax
6. Editions of Dungeons & Dragons
7. Dark Sun
8. Dungeons & Dragons campaign settings
9. Guildmasters’ Guide to Ravnica
10. List of Dungeons & Dragons rulebooks

Each of these tell a similar, but distinct story. But things get interesting when you treat Wikipedia as it’s own entity. Milne & Witten (2008) proposed a method of examining wikipedia not by looking at at whether A link to B, but by identifying which nodes link to both A and B. (I’ve used a modified method by Miccio et al., 2025, which averages inbound and outbound links). This graph shows connections with a relationship above .8.

What we see here is a different kind of story. There’s a core D&D cluster (with key players and game concepts), a video game cluster, a Hasbro (corporate/ other-game cluster), a fiction/genre cluster, and then a loose cluster that seems to capture things that were influenced by D&D (like non-D&D game makers, various kinds of awards, and publications/media). These communities are telling as a much more informative story.

But if look at the node metrics, we see something else, as well. It’s clearly telling us something about the content of the game rather than more superficial relationships. Tolkien is nowhere to be seen.

There’s a big open question (at least to me, who needs to do more reading on the topic) about what exactly a wiki link represents, and how the various biases of wiki influence the construction of the graphs.

But as a starting point for graphing freely available data, and making inferences, it strikes me there is much potential. I’ve used this to graph Joe Rogan (Surprise! He’s demonstrably a useful idiot for reactionary right-wing ghouls!), and Santa Claus (Real surprise… a much looser network with a strong food-themed community, and religion playing a very weak role overall). I have more sophisticated ideas I’m yet to pursue (and write about), but as a starting point, I think this is a useful proof of concept.

Cross-posted at theofficialwebsiteof.com

Fooling around with Blogs

Well, here I am.

I’ve attempted to outrun social media, and in so doing, have created a website that probably won’t even be indexed by google. Take that, the man!

I’m also a millennial, and blogs feel like what the internet should be.

I’ve been in this academic game for about 10 years now, and I think I might - in some limited contexts - have opinions on things that might be worth stating. Or at the very least, I’ve made enough specific mistakes that I can warn others not to do the same. (And if they’re worth their salt, they’ll go ahead and ignore me and make their own kinds of mistakes from which to learn). And, though luck has played a significant role, I do have some insights that are generalizable.

But there are two more immediate motives. First, I think it might be a nice way to exorcise various ideas that I would otherwise spend too much time on in a formal context. And second, I want to document ideas for works in progress (especially ideas that will ultimately not amount to much).

For example, I’ve been fooling around with graphs a lot. I’ve built a few tools using Claude* to do this. Some of these tools are modified approaches to collecting data from participants that produce graphs. Some are specific kinds of web-scrapers and graph-makers, that will let me ask specific kinds of questions of freely available data. The next post I make will be demonstrating some these methodologies. Some of it will turn into a publication, I hope. But most won’t. Things will be born here, and things will die.

But more seriously, I’m writing a book, and I want to create a record of that process. It’s tentatively called ‘Santa Claus and Religious Cognition’, and it’s a part of a Cambridge Elements serious on the ‘Psychology of Religion’. I’ve got until August next year to submit the manuscript, and while I’m not by-nature an 11th hour kind of guy, forcing a little bit of public accountability on myself can’t hurt. Probably.

I also make games. You can buy them! I tend to put the digital versions on rpg-trader, but physical copies of ‘The Last Job You’ll Ever Need’ can be found at here. (At least at the time of writing this blog in late 2026). Depending on what the world ‘professional’ means, and whether this blog remains ‘professional’, I will share thoughts about games, and games scholarship. (Though, having been floating around the space for a few years now, and having come to various conclusions… I may not commit those thoughts to text). We’ll see exactly how this comes out, but I do study imagination and reality beliefs, and I use games to do that.

So, here I am. Just fooling around with a blog.

* LLMs will get a post, for sure. I have thoughts, and while they’re probably not particularly insightful, they will stand as a published position statement on the topic.