four channels of goal inference

reward modeling and revealed-preference inverse reinforcement learning

Both reconstruct a principal's true objective from indirect evidence rather than accepting a declared objective at face value.

1 source videos