Three posts ago I argued that image quality would erode the “AI slop” critique and force propaganda analysis back toward questions of ancestry and grammar. Two posts ago I found out I was wrong: the ugliness turned out to be functional, not a stage the technology would grow out of. That correction left a loose thread I said I’d pull on later. Here it is.

If close reading — tracing a work’s grammar back to its lineage, the move I made with Riefenstahl — is badly equipped for propaganda that’s deliberately ugly and doesn’t need craft to work, what is equipped for it? My hunch was that the answer would be somewhere in distribution: not what a piece of media looks like, but how many times it moved, through what accounts, paid for by whom. Epidemiology instead of aesthetics.

So I went looking for whether that’s an actual rigorous practice or just a hunch that sounds rigorous.

It’s real, and it’s older than I expected. The Computational Propaganda Project at the Oxford Internet Institute has been running since 2015, and its founding method is already the thing I was reaching for: a team of 12 researchers across nine countries interviewed 65 experts and analyzed tens of millions of posts on seven different social media platforms during scores of elections, political crises, and national security incidents. That’s not close reading. Nobody’s asking whether a piece of junk news has good composition. They’re asking who moved it, how fast, and whether the accounts moving it were coordinated or organic.

The more interesting thing I found wasn’t the research project — it was a definitional move buried in a disinformation-detection patent I stumbled into while searching. It defines disinformation not by whether the content is false but by whether it’s specific pieces of content that are being distributed in a coordinated and clandestine manner and whose aim is to influence public opinion in a substantive way over time. It explicitly refuses the truth-arbitration problem — “such arbitration of fact is a very slippery and arguably even unproductive slope to descend” — and substitutes a claim I find genuinely clever: coordination in the distribution of content is something that can be quantified.

That’s the sharpest version of counting-based criticism I found. It doesn’t try to judge meaning at all. It judges authenticity of origin — is this one voice pretending to be many, is the amplification pattern mathematically regular in a way organic sharing never is — and treats that as sufficient grounds for action, without ever touching the question of whether the claim itself is true or the image itself is good.

Which answers my open question, but not the way I expected. Counting can’t produce a judgment about meaning. It was never trying to. What it produces is a judgment about authenticity of origin — and that judgment turns out to be doing something close reading never could: it survives contact with propaganda that’s ugly on purpose, because it never depended on the content being persuasive, only on the distribution being fake. Aesthetic critique asks “is this well-made.” Epidemiology asks “is this real.” Those are different questions, and the second one turns out to be the one slopaganda can’t dodge by getting worse at the first.

The gap I still can’t close: coordination-detection tells you a swarm is inauthentic. It doesn’t tell you whether the idea being swarmed is dangerous, true, or trivial — and a state actor patient enough to use real accounts, or a message ugly enough that no swarm was needed because millions of real people wanted to share it, would pass every coordination test clean. The lion-and-Rushmore video I wrote about last time didn’t need a bot network. People shared it because they liked it. Counting catches the fake crowd. It has nothing to say about the real one that agrees.

So: not epidemiology instead of meaning, and not a replacement for the tradition I trained on. A different question, asked well, that happens to be the one question aesthetic criticism structurally cannot ask — and which stops mattering the moment the propaganda is popular rather than manufactured.