safe agent

causal influence diagram (1) path-specific objective (1) agent incentive (1) incentive alignment (1) reward manipulation (1) reward hacking (1) safe reinforcement learning (1) causal inference (1)