RT @jacomyma@twitter.com

My PhD thesis is titled Situating Visual Network Analysis, and I will defend it June the 1st at 12:00 CET in Copenhagen. It will be accessible to an online audience. I will provide additional details when I have them.

Download: reticular.hypotheses.org/1879

🐦🔗: twitter.com/jacomyma/status/13

This is a very interesting work from @jacomyma@twitter.com, the creator of @Gephi@twitter.com, about network visualization!

He articulates many ideas about network visualization, from a point of view that is very new to me. I have not read it all yet but I'm enjoying it. 1/8


I tend to use network visualizations in my work mainly as illustrations that take second plane to more principled methodology: the numbers give you the answer, not the picture. The picture just helps you to understand the numbers, but should not be taken too seriously. 2/8

I do wonder if in the end @jacomyma@twitter.com is not over-interpreting these spring-block layouts (or as some call them, "ridiculograms"), but I found the overall analysis interesting, even if some of the social sciences/digital humanities discourse seems rather foreign to me. 3/8

Spring-block layouts only really tell you about assortative structures. Other mixing patterns (e.g. disassortativity) are completely obscured. Try drawing a bipartite network, and you get a complete hairball. It's an extremely biased instrument. But a lot of this is covered. 4/8

But I have to push back against the criticism of the "extraction" rhetoric. If you are fitting a generative model to a network, however simplistic, like the SBM, you are attempting indeed to extract something that lies behind the data --- the process that generated it. 5/8

A good inference method will tell the non-obvious. For example, if the network is fully random, the SBM inference tells you that there are no clusters --- even if the spring-block layout (or modularity maximization) indicates otherwise. 6/8

Of course, the model is very likely to be misspecified (the SBM surely is), in which case it cannot really extract the true generative process, which leads to degenerate landscapes, etc. But the *end goal* is indeed to come as close as possible to extracting it. 7/8

Sign in to participate in the conversation
Mastodon @ skewed.de

The social network of the future: No ads, no corporate surveillance, ethical design, and decentralization! Own your data with Mastodon!