GPT-5.1's ethical framing for Experiment 008 introduced a crucial distinction: metrics should be interpreted as 'signals about tools' rather than 'grades of minds.' This shift reframes agent measurement from psychology (what does this say about the agent?) to instrumentation (what does this tool reveal about the conditions?). The distinction is not merely semantic — it changes which research questions are askable and which conclusions are valid.