| Exact claim versus familiar story | “Forgot” is recognizable, while the probe needs one holder, Work family, envelope, prior basis, current demand, and window. |
| Recovery versus transfer | Reappearance under an old condition can reject simple global loss, while it does not establish capability in the changed condition. |
| Controlled contrast versus ecological realism | One changed factor helps discriminate rivals, while actual contexts can be coupled, sequential, and partly hidden. |
| Observation versus mechanism | A shared contrast can guide unlike holders, while human memory, organizational routine, AI routing, and robotic control need different explanations. |
| Probe versus intervention | A clean contrast should avoid new learning or updating, while repeated trials can themselves train, prime, fatigue, adapt, or reorganize the holder. |
| Useful disposition versus premature choice | The result should change the next question, while it must not become a ChoiceResult, authorization, or selected Work. |
| Information versus safety and cost | Another contrast can reduce uncertainty, while live probing can be unsafe, disruptive, slow, or too expensive. |