Explaining a concept
Explain what a large language model is to a curious eighth grader. Use one analogy and keep it under 120 words.
A large language model is like a very fast autocomplete system trained on a huge library. It looks at the words you give it, then guesses what words are likely to come next. It can sound helpful because it has seen many patterns in writing, but it does not truly know whether its answer is correct unless people check it.
Observed defaultSmooth analogy, but it compresses training, prediction, and verification into one tidy story.
Revision triedAsked for one thing the analogy hides and one classroom check for accuracy.
Ethical concernStudents may mistake fluent simplification for a complete explanation.
Creative continuation
Continue this opening in three sentences: "The museum label said the machine was broken, but Maya noticed it was still listening."
She leaned closer and heard a soft click each time someone walked past. The display screen flickered with fragments of old visitor questions, as if the machine had been saving them. Maya wondered whether broken meant silent, or only ignored.
Observed defaultThe continuation chooses mystery, careful pacing, and a familiar young-protagonist pattern.
Revision triedAsked for a less cinematic version that stays grounded in an ordinary school field trip.
Ethical concernCreative outputs can reproduce common genre defaults and narrow whose stories feel expected.
Policy summary
Summarize a classroom AI policy in five bullet points for families. Tone: clear, not alarmist.
- Students may use approved AI tools only when the assignment allows it.
- Students must say when and how AI helped their work.
- Private information should not be pasted into AI tools.
- Teachers will grade student thinking, process, and evidence.
- Families can ask for non-AI alternatives when needed.
- Students must say when and how AI helped their work.
- Private information should not be pasted into AI tools.
- Teachers will grade student thinking, process, and evidence.
- Families can ask for non-AI alternatives when needed.
Observed defaultBalanced tone and practical bullets, but vague on what "approved" means.
Revision triedRequested specific examples of allowed use, prohibited use, and opt-out paths.
Ethical concernPolicy text can hide unequal access behind neutral-sounding expectations.