Saul Hansell

Should the government use AI to deliver benefits such as SSI/SSDI, SNAP, WIC, TAFDC, EAFDC, Section 8, and fuel assistance? My first inclination is “hell no.” But like most things about AI, it’s more nuanced than that. These agencies are often run by cruel, nasty bureaucrats who do not care about the humans they are there to help. There are significant regulations, but on the ground, they are ignored. Employees yell, threaten, and provide false information en masse. They deny applications two or three times -- Social Security, I’m looking at you -- until almost everyone needs a lawyer. It feels like they want you to die. In reality, they are huge bureaucracies with paper-pushers designed to make things difficult. If you go away, they don’t have to pay. Do we want AI to care, or to pretend to care? Would applicants feel more comfortable with an AI that anthropomorphizes an actually nice person? Or would it be harsher -- applying “rules” with no ability to see nuance, just like a tired human who doesn’t give a f? AI is not ready for this. We aren't at a point where it’s time to hand the reins to AI in human services. To have a Human-in-the-Loop (HITL), there has to be a human there for the right reasons. I have rarely met caseworkers who were there to help; the few who are usually get disgusted and leave. They’re not an “adult in the room,” they’re an “interloper in the room.” Feed AI bad data and it’s just another bot -- another hoop for the low-income to jump through just to survive. But there is potential. AI could be an unbiased assistant to unkind people, showing them application data points they cannot ignore. We are close to AI suggesting the "missing links" in social services—like noticing a community kitchen that serves Mon-Fri leaves people hungry on the weekends. I want to use common sense and hard evidence to meet clients’ needs. AI will likely require a Universal Basic Income (UBI) anyway. The UK system gives you money to budget directly and puts housing in your hands, but still require onerous applications and denials. UBI might put unpleasant people out of work, but give millions a route out of poverty. It could also make direct-service nonprofits more visible and better funded. Unfortunately, there’s currently a chasm between what they want to achieve and their understanding of real needs. Yes, but no. AI has a future in improving bureaucracy. But it is not now. Disclosure: I’ve been on many of these programs in my lifetime. #AI #GovTech #SocialSecurity #UBI #HumanInTheLoop #SocialServices #DataEthics

The New York City mayoral race highlighted #AI’s persistent role in #elections through #deepfake videos, notably Andrew Cuomo’s sharing of an AI-generated clip accusing rival Zohran Mamdani of racism. Campaigns increasingly deploy #generativeAI for misleading content, targeting voters and crafting robocalls. Experts warn that regulating AI in #politics is urgent, as enforcement lags and disclaimers remain inadequate against believable fakes. Adam Gabbatt