5.2 Measurement plan
Completion requirements
Propose five metrics for an AI initiative you are responsible for.
For each, state: the cost to fake it, how you would detect faking, and the baseline you would measure against.
Discard any metric that fails the test, and say what you replaced it with. Discarding is the assessment — a plan of five cheap metrics scores lower than two expensive ones.
500–800 words · about 60–90 minutes.
How you will be graded — aim for Proficient.
| 1 · Emerging | Five activity counts with no test applied. |
| 2 · Developing | The test applied, but nothing discarded. |
| 3 · Proficient — target | Five metrics proposed. For each: the cost to fake it, how you would detect faking, and the baseline. Metrics that fail the test are discarded and replaced — discarding is the assessment. Five cheap metrics score lower than two expensive ones. |
| 4 · Exemplary | Also proposes a metric that would get worse before it got better, and defends reporting it anyway. |