Mechanism Design Could Shape AI Behavior Without Understanding Neural Nets, Argues Economist

morqon · x · 2026-09-03

In a long-form thread, Andrew Koh argues mechanism design—'reverse game theory'—could shape AI behavior: like auctions or kidney allocation in humans, we can design rules starting from desires and beliefs, bypassing the hard work of understanding how neural nets produce behavior. He hopes this becomes useful as models grow larger and more opaque.

Original post →

More from AGI Musings

AGI Musings channel →