Apprenticeship learning — borrows technique from → Inverse reinforcement learning (IRL)

Concrete Problems lists inverse reinforcement learning and apprenticeship learning side by side as the two techniques that can implement "use demonstrations" (concrete-problems, §"Use Demonstrations:", p. 15), but they target different objects. Inverse RL, in its classical form, tries to recover an explicit reward function (or a policy derived from one) that rationalizes the expert's trajectories. Apprenticeship learning, as introduced by Abbeel and Ng, does not aim to recover the true reward; it searches for a policy whose long-run feature expectations match the expert's, using an IRL-style optimization only as an internal subroutine toward that behavioral- matching goal. The distinction matters for safety: a technique that outputs a reward function can be inspected, reused, or checked for hacking vulnerabilities, while one that outputs only a matching policy cannot be interrogated for its implicit objective in the same way.