Framework components
Domains, lenses, and levels: what's the difference?
Each answers a different question. Domains: what to measure. Lenses: whether results are good enough to act on. Levels: the scope of use, who is accountable, and which decision applies.
The facts you collect: what went in (cost), how AI was used (use), and what changed (outcomes).
Produces: numbers and findings.
Five tests applied to those facts: quality, readiness, equity, governance and risk, and evidence. Each is a gate; failing one stops the use case.
Produces: pass, watch, or fail.
Where AI is used and measured: a person’s task, a team’s workflow, a unit’s services, or the institution’s portfolio. The level sets who is accountable and which decision applies.
Produces: a scope, an owner, and a decision. See the matrix
An advising assistant halves wait times (outcomes domain). But screen-reader users can’t book appointments, so the equity gate fails (lens). It runs campus-wide, so the institution decides (level): restrict it, despite the faster waits.
In short: measure with the domains, judge with the lenses, and decide at the right level.
PrintableThe framework on one pageOpen diagram →Open a component for its parts and where to read more.
Select a step for details.