This project began as a post about a year ago, in which we followed the provenance of the data from a study about misinformation until, at the bottom, we found Media Bias/Fact Check. Upon further examination, it turned out that Media Bias/Fact Check is just some guy, and upon him there rests a surprising amount of academic research. We've since turned this into a study with the following abstract:

Media Bias/Fact Check (MBFC) purports to quantify the bias, credibility, and factuality of reporting for roughly 10,000 media sources, and the resulting data is commonly used in misinformation research. In the present study, we show that MBFC's methodology does not meet basic standards of rigor for academic research. Despite its widespread prevalence, studies using MBFC rarely examine it carefully, often describing it in ways that contradict its "About" page, or treating it as authoritative despite MBFC's disclaimer that it is "not a tested scientific method... [but] a simple guide to the idea of a source's bias." We identified no papers that adequately describe MBFC as the opinions of a single person or critically engage with its methodology in order to justify proceeding with its use. We argue that MBFC's data is not neutral or accurate, but a computationally legible account of hegemony, a specious dataset for uncritical research that mistakes the familiarity of the concepts it quantifies with accuracy. Our study concludes with a call for academic researchers to stop using MBFC. MBFC's data quantifies the results of political processes, including campaigns to discredit the press, and presents them as simple facts about the world, thus reproducing the crisis misinformation scholarship exists to address.

In early 2024, shortly after writing a post on Maples et al.'s "Loneliness and suicide mitigation for students using GPT3-enabled chatbots," we also wrote a response to the journal. In it, we noted that Replika, the app they used for the study, was an AI companion app marketed for erotic role play. Maples et al. had failed to mention this little piece of context in their study, along with Maples's conflicts of interest.

As of November 2025, that response was finally published, along with their reply. Despite apparently taking a year to write, their reply is a mere 319 words and dodges the substance of our critique. In this second reply, we go through their response point by point and, in the process, show how both Maples and Luka (the company that owns Replika) have used the study in their advertising and public communications, greatly exaggerating and generalizing its already dubious findings as it suits them.

Since 2016, the number of papers on mis/disinformation has skyrocketed. It's hard to imagine another phenomenon for which there exists such a complete dataset, but this field must label things as true, biased, etc., sticky concepts to define at scale. How can you label millions of tweets as true/false?

If we pick a paper and follow the provenance of its data, I think that we can learn about not just this field, but the role of data and computation in society.

cross-posted from: https://lemm.ee/post/56052351

Dávila's "Blockchain Radicals" argues that the left ought to embrace blockchain. Here's my 2 part review. The first critiques the book's approach to argumentation, and the second examines Dávila's own Breadchain Cooperative.

This is my longest post yet because the theory the book presents is palatable to developers. It does to political theory what tech people always do: Confidently assume their skills apply in a field they don't bother to understand. The consequences are predictable. This, then, is an intervention directed at that mode of thinking, an examination of how bad theory leads to bad practice, and, most importantly, an attempt to stop would-be activists from getting caught up in this mess.

tl;dr Breadchain's use of the term "cooperative" is fraudulent, and it is, structurally, a grift, whatever his intentions might be.

Though wrapped in the aesthetic of science, this paper is a pure expression of the AI hype's ideology, including its reliance on invisible, alienated labor. Its data was manufactured to spec to support the authors' pre-existing beliefs, and its conclusions are nothing but a re-articulation of their arrogance and ideological impoverishment.

The book "#HashtagActivism" is a robust and thorough defense of its namesake practice. It argues that Twitter disintermediated public discourse, analyzing networks of user interactions in that context. But the book overlooks that Twitter is actually a heavy-handed intermediary. Twitter imposes strict requirements on content, like a character limit, and controls who sees what and in what context. Reintroducing Twitter as the medium and reinterpreting the analysis exposes serious flaws. Similarly, their defense of hashtag activism relies almost exclusively on Twitter engagement data, but offers no theory of change stemming from that engagement. By reexamining their evidence, I argue that hashtag activism is not just ineffective, but its institutional dynamics are structurally conservative and inherently anti-democratic.

Regulating tech is hard, in part because computers can do so many things. This makes them useful but also complicated. Companies hide in that complexity, rendering undesirable behavior illegible to regulation: Regulating tech becomes regulating unlicensed taxis, mass surveillance, illegal hotels, social media, etc.

If we actually want accountable tech, I argue that we should focus on the tech itself, not its downstream consequences. Here's my (non-environmental) case for rationing computation.

Until recently, platforms like Tinder and Uber couldn't exist. They need the intimate data that only mobile devices can provide, which they use to mediate human relationships. They never own anything. In some ways, this simplifies their task, because owning things is hard, but human activities are complicated, making them illegible to computers. As tech companies become more powerful and push deeper into our lives, here's a post about that tension and its consequences.

I've seen a few articles like this one from Futurism: "CEOs Could Easily Be Replaced With AI, Experts Argue." I totally get the appeal, but these articles are more anti-labor than anti-CEO. Because CEOs can't actually be disciplined with threats of automation, these articles further entrench an inherently anti-labor logic, telling readers that losing our livelihoods to automation is part of some natural order, rather than the result of political decisions that benefit capital.

Lots of skeptics are writing lots of good things about the AI hype, but so far, I've encountered relatively few attempts to explain why it's happening at all. Here's my contribution, mostly based Philp Agre's work on the (so-called) internet revolution, which focuses less on the capabilities of the tech itself, as most in mainstream did (and still do), but on the role of a new technology in the ever-present and continuous renegotiation of power within human institutions.

The video opens with Rober standing in front of a fancy-looking box, saying:

Hiding inside this box is an absolute marvel of engineering you might just find protecting you the next time you're at a public event that's got a lot of people.

When he says "protecting you," the video momentarily cuts to stock footage of a packed sports stadium, the first of many "war on terror"-coded editorial decisions, before returning to the box, which opens and releases a drone. This is no ordinary drone, he says, but a particularly heavy and fast drone, designed to smash "bad guy drones trying to do bad guy things." He explains how "it's only a matter of time" before these bad guys' drones attack infrastructure "or worse," cutting to a photo of a stadium for the third time in just 30 seconds.

In "If We Burn," Vincent Bevins recaps the mass protests of the 2010s. He argues that they're communicative acts, but power has no way of negotiating with or interpreting them. They're "illegible."

Here's a "yes and" to Bevins. I argue that social media companies have a detailed map of all protesters' connections, communications, topics of interests, locations, etc., such that, to them, there has never been a more legible form of social organization, giving them too much power over ostensibly leaderless movements.

I also want to plug Bevins's book, independently of my post. It's extremely well researched. For many of the things that he describes, he was there, and he productively challenges many core values of the movements in which I and any others probably reading this have participated.