# Learning, understanding, and oversight exercise

Purpose: investigate whether a workflow helps people perform the task while retaining the understanding needed to check and recover it. This is a proposed work-design exercise, not a diagnostic instrument or employee ranking.

For an ongoing coaching and progression plan, pair this exercise with [User development plan](09-user-development-plan.md) and `user-progress-log.csv`. This document tests a learning arrangement; the plan tracks the next practice step for one capability.

**Learning objective / task / facilitator / participant:** [fill in]

**Prior familiarity / accessibility needs / permitted assistance:** [fill in]

**Announced immediate check / delayed check date / access to results:** [fill in]

## Design the comparison

Compare task-oriented assistance with an explanation-and-practice approach, or compare either with the current process. Match task difficulty and experience where possible. Decide the rubric in advance and describe the conditions to participants. Do not claim causality from an uncontrolled comparison.

For the explanation-and-practice condition, ask the participant to make an initial prediction, use assistance, and explain the final choice in their own words. Do not require disclosure of a model's hidden reasoning; assess the participant's explanation and observable evidence.

Keep assisted operating performance separate from any unaided understanding check. Assistance should be allowed where real work requires it; an unaided check answers a different question about retained knowledge.

## Exercise and descriptive rubric

| Prompt | Evidence to look for | Result |
| --- | --- | --- |
| Explain a consequential choice. | Names the objective, source, tradeoff, and an assumption. | Demonstrated / partial / not demonstrated / not observed |
| Find a planted error in fictional output. | Identifies the error and verifies the correction against a reference. | [fill in] |
| Predict an unfamiliar exception. | Explains expected behavior and when to stop or escalate. | [fill in] |
| Recover a failed test change. | Identifies affected state, contains the issue, and verifies recovery. | [fill in] |
| Hand the work to another operator. | The recipient can locate evidence and explain the remaining uncertainty. | [fill in] |

Record the observation and evidence for each result. Do not sum these labels into a universal competence score. Not observed means the exercise did not establish the capability; it is different from demonstrated failure.

At a pre-agreed later date, use a different but related task to examine transfer. Record time since training and practice in between. If follow-up is missed, retention remains unmeasured.

**Immediate observations / output quality / effort:** [fill in]

**Delayed observations / follow-up coverage / alternative explanations:** [fill in]

**Training or workflow change to try / owner / next check:** [fill in]

Evidence that assistance affects immediate learning in a particular task does not establish permanent cognitive decline. Evidence of workplace gains does not guarantee transfer to this task. Use the exercise to improve the arrangement and describe its limits.