pluralistic.net web signal

Doctorow: DOGE turns federal workers into AI blame-takers

TL;DR

  • Cory Doctorow argues the real AI market is not chatbots but 'decision support' systems for medical, bail, resume, and loan calls.
  • He cites a PCMag report that Musk's DOGE is accelerating AI across Medicare, food stamps, VA benefits, visas, indictments, bail, and parole.
  • The lone 'human in the loop' becomes a 'moral crumple zone,' present to absorb blame for machine errors rather than prevent them.

Cory Doctorow's Pluralistic post asbestos in the walls argues that the interesting AI market is not the one grabbing headlines. Chatbots and image generators, he writes, sell into commercial markets that are economically thin compared with their hype. The real money is in 'decision support,' the software that automates high-stakes calls like medical diagnosis, bail and parole decisions, resume screening, and loan applications.

His central metaphor is the reverse-centaur. The optimistic story about AI is a human clinician who gets an AI second opinion. The version being sold to buyers, he says, is the opposite: 'Fire all but one of your radiologists and then put that poor bastard to work reviewing the judgments our robot makes at machine scale.' One person cannot maintain vigilance over a firehose of machine outputs, and the humans who remain are structurally set up to fail.

The immediate hook is a PCMag report that Doctorow cites on Elon Musk's DOGE accelerating an AI rollout inside the federal workforce, touching decisions on Medicare, food stamps, VA benefits, visas, indictments, bail, and parole. His framing is blunt: 'Musk is transforming key government functions into high-speed error-generating machines whose human minders are only the payroll to take the fall for the coming tsunami of robot fuckups.' That lone reviewer, in his telling, is what a cited academic paper calls a moral crumple zone. 'The role of this "human in the loop" isn\'t to prevent errors,' Doctorow writes. 'That human is there to be blamed for errors.' Two experts in our Who's Who directory circulated the source link.

He layers a technical point on top about feedback loops. 'AIs want the future to be like the past, and AIs make the future like the past.' If training data carries historical bias the model reproduces it, and automated decisions that feed back into training compound it. The image he lands on is that a federal build-out along these lines 'is the equivalent to filling the American government\'s walls with asbestos, turning agencies into hazmat zones that we can\'t touch without causing thousands to sicken and die.'

What the post does not do is name the specific DOGE contracts or vendors, put a dollar figure on any of it, or describe how a Medicare or SNAP claimant would appeal an automated denial. It is an opinion essay leaning on a single PCMag scoop, not a procurement investigation. But as a frame for who ends up carrying the liability when an AI-in-the-loop program goes wrong inside a benefits agency, it is the argument buyers of these systems, and the reporters watching them, will want in their heads before the first wrongful-denial lawsuits land.

Shared on Bluesky by 2 AI experts