Principal-Agent Problem
The principal-agent problem arises when an agent is rewarded for a measurable proxy rather than the outcome the principal actually values. Redirecting payment toward that outcome can transform behavior without increasing the total budget.
Ship captains once delivered voyages successfully while half their prisoners died. When payment shifted from completing the voyage to delivering each prisoner alive, survival jumped from 50% to 98%.
E1The contract defines the real job
A principal delegates work but cannot directly control every decision the agent makes. The agent therefore optimizes whatever determines payment: here, completing a voyage rather than preserving lives. Deaths mattered morally to the principal, but not economically to the captain. Paying per living prisoner made survival part of the captain's own result. This is incentive alignment in its sharpest form: the same money buys radically different behavior when attached to a different target.
E1Where it shows up
From transport to survival
The captains did not need a larger budget or a new stated mission. They needed the outcome previously treated as someone else's concern—prisoner survival—to determine their compensation.
E1A target is still only a target
The lesson is not that every important outcome should mechanically determine pay. This case had an unusually clear, countable result: prisoners arrived alive or did not. When quality is ambiguous, delayed, or easy to manipulate, a new metric may simply create a new perverse incentive.
E1Rewrite one payment trigger
Take one delegated task and write down two things: the result you actually want and the event that currently releases payment or earns credit. If they differ, redesign one milestone so compensation depends on verified delivery of the real result—not merely completion of the agent's activity.
E1Episodes that teach this
-
You're Not Lazy - You're Just Rewarded Wrong - Future IQ
· explained at 3:33
9,886 views
Ship captains were paid for the voyage, not for keeping prisoners alive; when paid per live prisoner delivered, survival rose from 50% to 98%.