# Aisha's AIRS Cohorts — Schema

Flat, aggregate-only teaching dataset in `aisha-airs-cohorts.csv`.

## Columns

| Column | Type | Description |
| --- | --- | --- |
| `cohort_id` | integer | Sequential cohort number (1 = earliest, 6 = most recent). |
| `semester` | text | Semester the cohort took the course (e.g. `Fall 2023`, `Spring 2024`). |
| `start_score` | number | Cohort mean AIRS score at the start of the course (0–100). |
| `end_score` | number | Cohort mean AIRS score at the end of the course (0–100). |
| `gain` | number | `end_score` minus `start_score`; positive means the course moved the cohort. |

## Row inventory

- 6 cohort rows (Fall 2023 through Spring 2026), one per semester.

Total: 6 rows.

## Deliberate design choices

Unlike the four cross-chapter teaching datasets (CloudRevenue, M365Marketing, SupportInsights, EnterpriseGovernance), this dataset has no injected null-value or naming-inconsistency imperfections. Its purpose is to reproduce Chapter 6’s Aisha figure exactly, and the figure's defects live in the *palette-type mismatch and encoding choices*, not in the data.

The one designed-in tension is the gain gradient across cohorts: Cohort 6 shows a materially larger gain than Cohort 1 (`+24` vs `+11`, roughly `2.2x`). That contrast is the Big Idea Aisha's figure is trying to communicate; a chart that hides it (as her brand-palette version does) is failing its audience.

## License

Released for teaching use alongside *The Defensible Decision*. Use freely.