Workshops
Sessions on judging what an AI product writes: who decides what good looks like, how to tell a cue from an answer, and what a quality number can and cannot claim.
-
Whose “Good Enough”?
Three 75-minute sessions for a product team: the literacy reviewer, the engineers who write the quiz generator, and the manager who decides.
Each person writes their own standard for a quiz item before hearing anyone else's. The group then applies five answer-choice rules to real items from the product's test stories, works out what reading load means at each grade band, and settles a disagreement between the human rater and the model judge with a decision record that names an owner. Judging is done blind, with a giveaway list so each person can mark why an answer stood out without the story.
-
Score It Yourself: How Reliable Is the AI That Grades Your AI?
Women Who Code Leadership Summit, New York, 8 October 2026. Thirty minutes for leaders who approve AI work and the engineers who build it.
Five ways a quality score misleads, three questions to ask about any AI number, and an exercise you can do from your seat: answer comprehension questions without the stories they belong to, then see how the AI evaluators did on the same items.
More series for the same team are written and will appear here as they are run: a six-session series on turning a concern into evidence, one on reading a change as a claim, and an incident learning lab.