Amina’s morning began with a regulatory briefing she didn’t fully understand and a croissant she fully appreciated.
She was a secondary school teacher in Geneva, and today was her day to serve on the Citizens’ AI Oversight Panel—a rotating civic duty, like jury service, introduced two years ago. Twenty residents, selected by lottery, spent a day reviewing how the city’s AI systems were being used: traffic management, school assignment algorithms, the new healthcare triage system that had replaced the emergency room queue.
The panel met in a government building with good coffee and terrible chairs. A facilitator walked them through the agenda. Today’s focus: the school assignment algorithm had been flagging immigrant children for “additional support” at rates that some parents considered discriminatory and others considered essential. The data was ambiguous. The algorithm’s developers said it was working as intended. The parents said intentions were not the point.
Amina listened. She asked questions. She thought about her own students—twelve-year-olds who used AI tutors at home and arrived at school knowing things she hadn’t taught them yet, but who couldn’t explain how they knew. She thought about the girl last week who’d submitted an essay that was too good, and the conversation they’d had about what “your own work” meant now.
The panel deliberated. They recommended three changes to the algorithm’s parameters. The recommendations would go to the city council. The council would probably accept two of them. The process was slow, imperfect, and entirely human.
Amina walked home through the old city. A delivery drone passed overhead. A tram drove itself past a café where a man was arguing with his phone about a parking ticket that had been issued by an automated system that was, the man insisted, wrong. The phone was patient.
It was not a revolution. It was not a crisis. It was a Tuesday in a city that was trying, with limited success and genuine effort, to govern a technology it did not fully understand. The croissant had been better than the governance, but the governance was getting better. Slowly. Humanly. Imperfectly.