Judgment strand · Add-on to From Sand to Agents

Students who can check the machine.

Machines give confident answers, and some of them are wrong. The Judgment strand adds one scored task to each grade band. Students say exactly what they want, check what comes back, and explain how they know. Some items have a mistake and some do not, so “this one is right” is a real answer and earns full credit.

What a school gets

  • One printable task kit per band: student packet, one-page teacher page with answer key, scoring sheet, family-letter insert
  • A 4-level rubric on three criteria; about 3 minutes to score one student
  • Runs fully unplugged in one class period. No student devices, accounts, or data
  • Attaches to units the band already teaches; no new course

How machine use is handled

  • K-5: no student AI use. Students check printed machine “guesses”
  • 6-8: printed outputs; an optional teacher-run demo for the whole class
  • 9-12: printed assistant outputs by default; student-run AI only where school policy allows, scored the same way
6-8The Machine Underneath

verify against sources; first delegation · Attaches to Unit 2: What Machine Learning Really Is; Unit 4: Bias, Data Quality, and Evaluation

What the student does

Students check three short AI-written answers, claim by claim, against a printed source sheet and keep a handwritten verification log.

What the teacher scores

The teacher scores the log and a 2-minute spoken check on a 3-criterion rubric in about 3 minutes per student.

Same VA CS SOL codes as the attached units: 6.DA.4, 8.DA.2

Pilot terms: free semester pilot · Licenses: pricing · Full arc: scope & sequence

Judgment strand: check the machine · Schoolhouse