On the Feasibility of Learning, Rather than Assuming, Human Biases for Reward Inference

OntheFeasibilityofLearning,RatherthanAssuming,HumanBiasesforRewardInferenceRohinShah1NoahGundotra1PieterAbbeel1AncaD.Dragan1Abstractp(as)µebQ(s,a;r)w!Ourgoalisforagentstooptimizetherightre-wwardfu...