When can a reward metric miss its goal?

A reward metric can miss its goal when the measured quantity differs from the outcome that matters. A metric is the measure used to award points or judge performance. The fictional card-making goal is more verified usable cards, but one possible reward counts every submitted card instead.

The two strategy records separate those measures. Fast submits ten cards, of which four are verified usable. Careful submits six, all six verified. Fast therefore leads on submissions, while Careful leads on the stated goal. The names label records, not judgments about anyone’s character.

At two points per submitted card, Fast earns twenty points and Careful twelve. A chooser assumed to care only about these points would select Fast. That makes the rule misaligned for the two records: its reward winner produces fewer verified usable cards than the alternative.

At two points per verified usable card, the rewards change to eight for Fast and twelve for Careful. The records and goal have not changed; the measured basis of reward has. Careful now wins both comparisons, so this alternative metric is aligned within the toy table.

Alignment is conditional, not a universal prediction about human motivation. Real people may value other outcomes, and verification can be imperfect or costly. The table explains why the relationship between a goal and its measure matters without claiming one reward rule will always produce the intended behavior.