When a program head asks why their plan fell short on an outcome, there are two kinds of answer. One is an impression: the reviewer felt the measure was weak. The other is a reference: the plan names the capstone project as the measure for this outcome, but does not say how the project is assessed or what counts as success. The second answer can be checked, discussed and acted on. The first can only be accepted or resented.
ProverbAI is built so that every finding is the second kind.
What a finding contains
For everything the review considers, ProverbAI's review records a finding, the passage in the plan that the finding rests on, and where that passage appears. If the plan does not address something, the finding says so plainly, and nothing is cited, because there is nothing to cite. A reviewer then confirms or adjusts each finding before anything reaches the program.
How findings become a program's result is part of ProverbAI's own review methodology, and that methodology stays with us. What institutions see is what they need to act on: each finding, the evidence behind it, and the feedback for the program.
Why the evidence matters
It keeps review honest
A reviewer confirms every result. With the evidence in view, reviewing means reading the cited passage and judging whether it supports the finding. Without it, the reviewer has to repeat the search ProverbAI has already done, and the temptation is to skim and approve. Evidence turns a review into a comparison: does this passage say what the finding claims?
It makes disagreement specific
Two careful readers can disagree about a plan. When both are looking at the same passage, the disagreement is about that passage, and it can be settled or recorded. When they are comparing impressions, it usually cannot.
It makes feedback usable
Feedback that refers to a specific section of the plan tells a program head where to start. "Strengthen your measures" does not.
It keeps the reasoning after people move on
Next year, a new reviewer or a new program head may want to know why a finding was made. The evidence is still there, attached to the finding, long after the person who first read the plan has moved on.
What evidence does not prove
A citation shows where a finding came from. It does not show that the finding is right. A passage can be cited accurately and still be misread, and the decisive sentence may sit elsewhere in the plan. That is why a reviewer confirms every result, and why a reviewer who disagrees can change a finding, provided they record the reason.
We are evaluating ProverbAI against trained reviewers on real plans, and will publish results when a blind test gives us evidence we can stand behind. Until then, we make no claim about how often its findings are right. The evidence is there so that nobody has to take them on trust.
Making plans easy to cite
Citations work best when the plan can be read exactly. Three habits help, whoever is doing the reading:
- Submit a digital document exported from the original, not a scan of a printout, so the text can be read and cited word for word.
- Keep everything in one document, including the curriculum map, the instruments and the appendix, so nothing the review needs sits behind a link.
- Use page numbers and clear section headings, so a program head reading the feedback can find the same place quickly.
