← Evaluating AI Output
Tomorrow Ready · Evaluating AI Output

The Evaluation Gate for Learning Languages, Years 7 to 8

Subject adaptation · Years 7–8 · Learning Languages · Field-Based STEM · Tony Jones

AI translation tools produce grammatically correct language that no one would actually say. The Evaluation Gate requires students to name criteria and compare options before any AI output is accepted as their own.
1Name
Candidates
2State
a Criterion
3Compare
Against It
4Justify
the Choice
The Strategy
  1. Present the language task. Before any tool use, students write two candidate phrasings or translations for the key expression or task element.
  2. Students name one criterion that matters for this specific task: grammatical accuracy, natural register, cultural appropriateness, or audience fit.
  3. Students compare their two candidates against that criterion and select one.
  4. They write two to three sentences justifying the choice, including what the unchosen option would have got wrong for this particular context and audience.
  5. Collect the gate record before any final drafting or tool use begins. It travels with submitted work.
Year-Band Practice

Years 7–8: Beginners

Focus on register. Would a native speaker of this age say this to a friend, or to a kaiako? That distinction is a cultural judgement no AI can make for the student. Students must name it before they choose.

Years 7–8: Continuing

Add audience fit as a criterion. The same message in te reo Māori, te Gagana Samoa, or French carries different cultural weight depending on who receives it and in what context. AI output will not carry this awareness.

AI Use Note

If students generate both candidate phrasings using AI, the gate has not worked. At least one candidate must come from the student's own current knowledge of the language.

Implementation

Decision Checkpoint

The Evaluation Gate is most powerful in language tasks involving register, tone, or cultural context. AI produces fluent but contextually hollow output in these areas. Students who have named a criterion can explain their choice; students who accepted AI output cannot.

Teacher Judgement Note

If students cannot explain why one phrasing is more culturally appropriate than another for this context and audience, they have not yet engaged with the language learning the task was designed to produce.

Related Frameworks

Evaluation Gate · Context Triage · Comparison Before Conclusion

Tony Jones · Founder, Field-Based STEM · Tomorrow Ready Resources · Free to use and share