What can a score hide about the people behind it?

A score summarizes a record, so its meaning depends on what was counted. A completion rate divides completed tasks by attempted tasks. When easy and hard tasks are combined, each group contributes according to the number of attempts it contains. Different mixes can produce a surprising overall comparison.

In the pictured fictional records, B completes 95% of easy tasks and 20% of hard tasks, above A’s 90% and 10%. Yet A completes 82% overall while B completes 35%. A has many more easy tasks in its mix: 90 of its 100 attempts, compared with 20 of B’s 100. The combined rate gives those easier tasks greater weight in A’s score.

The chosen measure can affect behavior. If a toy reward pays only for submission count, increasing submissions may earn more points without improving the intended goal of verified work. That result follows from the stated reward rules; real people may respond in many different ways.

Wording can emphasize one side of a record: “80 successes” and “20 failures” may describe the same 100 attempts. Timing can also tempt a stronger conclusion than the evidence supports. An increase after a reward begins does not isolate the reward’s effect without a suitable comparison and attention to other changes. A useful score needs its denominator, context and limits.