Visualizing Divinatory Networks
Part 5 of this week's series on algorithms/analytics/prediction/divination . . . .
The main story for today is something like “How Nodus Labs used the sophisticated process of ‘network visualization’ to analyze Tarot structures.”
On a Lighter Note
Researching this week’s series, I came across The Tarot Cards of Tech—”a tool to inspire important conversations around the true impact of technology and the products we design.” This set of downloadable cards is meant for use in team discussions, to raise questions that might not otherwise come up.
Obviously, they are using the word “Tarot” to mean something like “illustrated pack of cards.” And probably they are taking advantage of the increased interest in Tarot. For an excellent overview of that trend, check out this story in the Washington Post.
But Tarot Tech reminds me of other decks that have been promoted as “creativity packs.” The first one I know of was the Creative Whack Pack, based on the 1973 book A Whack on the Side of the Head: How You Can Be More Creative, by Roger Von Oech.
And as I’m behind on January Fools, I’ll appropriate the Whack Pack’s “Foolish” jester:
Since then, several other creativity decks have come along—and one of the most recent examples describes itself as “a synthesis of Design, Tarot, Numerology and Gestalt Psychology.” The Intùiti deck follows Tarot structures, dividing 78 cards into “primary” and “secondary” groups.
The primary group offers a high-energy abstraction of the Tarot trumps:
The other 56 cards are apparently divided into four suits, but I didn’t have time to figure out more than that.
You can get a three-card Intuiti reading—which I found quite interesting . . .
These three cards represent (left to right) Hermit, Wheel of Fortune, and —The Fool. So it appears I’ve “fortuitously” caught up on January Fools!
Now, on to Nodus Labs, which “does research in communities, communication, and social artefacts through the frameworks of complexity theory and network science.” Their key product, InfraNodus, utilizes “algorithms for text network analysis and visualization that offer novel ways of doing text mining and topic modeling, as well as discourse and narrative structure analysis.”
The easiest way to describe InfraNodus would be to call it a very (very) fancy word cloud generator. You can see the visual output, and even play with it, in this analysis of movie genre descriptions.
But Nodus Labs also utilizes other types of analysis to discover relationships and create network visualizations. Apparently two members of the Nodus team had an interest in Tarot, which they explained this way:
Tarot is a very efficient mechanism of reading, which allows access to the whole story through randomly choosing only a few of its parts. We were interested to study the features of Tarot as a system that make it such a powerful reading device.
The report of their study, The Divination Network of Tarot, is written quite briefly, and there’s not a lot of information about the processes they used. Here’s the gist of it:
In order to study the structure of Tarot as a narrative tool, we needed to model it as a network. We represented the 22 main cards (from 0 to 21) as the nodes and their relations as the edges. We created several networks for different types of relations in order to be able to study them separately and to also compare information from different sources.
By whatever means they used, Dmitry Paranyushkin and Colin Johnco came up with several interesting conclusions. I’ll try to illuminate three of them.
First: I love anything that has to do with Tarot/numbers/patterns, so I’ll start here.
The authors assert that since the highest number on the cards is the 21, any two cards are connected if they add up to that “perfect” number. For example, 6 + 15 = 21 . . .
That’s the only example they provide—but it doesn’t take long to figure out the other combinations:
I found some of these combinations spot-on, while others seemed odd. But altogether, I thought this was a thought-provoking exercise.
Second: The modelers created a graph (never mind how!) of the Major Arcana relationships. And . . .
The main property of the graph we obtained was its very specific community structure. Based on a modularity algorithm, we found 3 distinct “communities”: cards that are more densely linked together [with each other] than [they are] with the rest of the network. Those communities are actually topics inside the Tarot narrative, which we separated into 5 groups.
So—here’s their illustration, followed by my attempt at clarification:
Here again—I find these groupings not exactly obvious, but worth thinking about.
Third: The authors came to a conclusion that four of the trumps have the “highest betweenness centrality.” That’s a measure of which nodes serve as bridges from one part of a graph to another. The trumps cited are Empress, Moon, Emperor, and Star.
Since I don’t have the data, I can’t say much about this conclusion. But! The authors suggest those four cards will appear more often than others when reading Tarot—and when they appear, will represent meaningful connections between different topics.
If that were true, statistically, I don’t think we would ever notice it. And I don’t think we could ever prove it, since there are so many variables involved.
But I’m intrigued by the idea of viewing some cards as topic-connectors.
We’re almost there! But it’s important to add that in 2016, Dmitry Paranyushkin and Colin Johnco used their Tarot analysis to build the Divinatorium—an interactive set of “abstracted divinatory narratives. I’m not sure what to think of it, and it seems to be unfinished. But I did find there a more elaborate version of the Connections Graph:
Some of the words that show up on the graph are unexpected—but fun to explore.
And speaking of fun, I’ve really enjoyed writing this week’s series. Thanks for reading!
Sunday’s newsletter will wrap up the month of January. And will include an expanded preview of Tarot | In Four Dimensions, which gets started next week.
Warmest regards, Cynthia