The PreCrime system doesn't work because precogs see the future—it works because precogs are impossible to replace, and Spielberg made that limitation architectural.
Watch the moment Anderton realizes the precogs are being gamed: they're not producing errors, they're producing *dissent*.
One precog sees murder. Two don't.
By 2002, algorithmic prediction already outpaced human intuition in finance, medicine, and targeting. But Spielberg couldn't imagine a PreCrime that didn't need to *see* a human face making the call — because the precogs exist so someone can be blamed. The moment you automate prediction, you've automated culpability, and no government in 2002 could stomach that.
The precogs were always a mercy.
”Spielberg's film warned us about surveillance. What it actually predicted was the horror of *automating accountability*—of building prediction systems so efficient that the only remaining human element is the person who implements the system's verdict. We chose the machine.
Read Rashida Richardson's essay 'Ocked In: Abolitionism, Police Technology and a World Without Cages' (2022) to see how PreCrime's algorithmic descendants actually function in American cities—and how removing the precog layer made the system harder to resist, not easier.
Jan Gehl's interview in the DVD commentary reveals Spielberg rejected early concept art that showed fully automated PreCrime headquarters, explicitly asking the design team to add the precog tank as a visual and mechanical anchor—he wanted audiences to *see* the human element so they'd understand what automation erases.