Why This Matters Now
Clarity beats cleverness—if a student cannot restate the task, the prompt is too ornate. Document your playbooks; new colleagues and substitute teachers should onboard in one afternoon.
Start with outcomes, not tools; prompts should map to your learning objectives and Bloom levels. Retrieval practice still wins—space it over days and mix in short, targeted hints from the model. Teacher time is precious; automate the repeatable, keep judgment and pastoral care human.
A Practical Framework
Honor privacy: minimize personal data, use district accounts, and rotate identifiers in exports. Teacher time is precious; automate the repeatable, keep judgment and pastoral care human. Start with outcomes, not tools; prompts should map to your learning objectives and Bloom levels.
For accessibility, provide multi‑modal options: text, audio, and captioned video instructions. Clarity beats cleverness—if a student cannot restate the task, the prompt is too ornate.
Retrieval practice still wins—space it over days and mix in short, targeted hints from the model. Always show a model answer and the rubric; feedback becomes legible and less surprising. Cold prompts underperform; prime with prior knowledge and short exemplars before free response.
Use chain-of-thought sparingly and never grade it; grade the final work against transparent criteria. Guard rails: forbid disallowed sources, cite where appropriate, and log versions for academic honesty.
Prompts that Work (Examples)
Teacher time is precious; automate the repeatable, keep judgment and pastoral care human. Always show a model answer and the rubric; feedback becomes legible and less surprising. For accessibility, provide multi‑modal options: text, audio, and captioned video instructions.
- Socratic: “Ask me one question at a time to test my understanding of photosynthesis. Increase difficulty as I succeed.”
- Rubric-driven feedback: “Score this essay on clarity, evidence, and structure (1–4 each). Return one strength and one next step.”
- UDL option: “Offer three representations of this concept: a 100‑word summary, a labeled diagram description, and a real‑world analogy.”
Guard rails: forbid disallowed sources, cite where appropriate, and log versions for academic honesty. Start with outcomes, not tools; prompts should map to your learning objectives and Bloom levels. AI is not a shortcut to learning; it is a mirror that requires better questions and stronger rubrics.
Cold prompts underperform; prime with prior knowledge and short exemplars before free response. Document your playbooks; new colleagues and substitute teachers should onboard in one afternoon. Retrieval practice still wins—space it over days and mix in short, targeted hints from the model.
Assessment & Academic Integrity
Document your playbooks; new colleagues and substitute teachers should onboard in one afternoon. Always show a model answer and the rubric; feedback becomes legible and less surprising.
For accessibility, provide multi‑modal options: text, audio, and captioned video instructions. Use chain-of-thought sparingly and never grade it; grade the final work against transparent criteria.
Teacher time is precious; automate the repeatable, keep judgment and pastoral care human. Clarity beats cleverness—if a student cannot restate the task, the prompt is too ornate.
Rollout in 2 Weeks
Guard rails: forbid disallowed sources, cite where appropriate, and log versions for academic honesty. Teacher time is precious; automate the repeatable, keep judgment and pastoral care human.
Document your playbooks; new colleagues and substitute teachers should onboard in one afternoon. Always show a model answer and the rubric; feedback becomes legible and less surprising.
For accessibility, provide multi‑modal options: text, audio, and captioned video instructions. AI is not a shortcut to learning; it is a mirror that requires better questions and stronger rubrics.
Pitfalls & Safeguards
Cold prompts underperform; prime with prior knowledge and short exemplars before free response. Teacher time is precious; automate the repeatable, keep judgment and pastoral care human. Retrieval practice still wins—space it over days and mix in short, tar geted hints from the model.
Start with outcomes, not tools; prompts should map to your learning objectives and Bloom levels. Document your playbooks; new colleagues and substitute teachers should onboard in one afternoon.
Use chain-of-thought sparingly and never grade it; grade the final work against transparent criteria. Guard rails: forbid disallowed sources, cite where appropriate, and log versions for academic honesty.
Always show a model answer and the rubric; feedback becomes legible and less surprising. For accessibility, provide multi‑modal options: text, audio, and captioned video instructions.
What to Measure
Start with outcomes, not tools; prompts should map to your learning objectives and Bloom levels. Teacher time is precious; automate the repeatable, keep judgment and pastoral care human.
Guard rails: forbid disallowed sources, cite where appropriate, and log versions for academic honesty. Honor privacy: minimize personal data, use district accounts, and rotate identifiers in exports. AI is not a shortcut to learning; it is a mirror that requires better questions and stronger rubrics.
Document your playbooks; new colleagues and substitute teachers should onboard in one afternoon. Cold prompts underperform; prime with prior knowledge and short exemplars before free response. Always show a model answer and the rubric; feedback becomes legible and less surprising.
Case Notes
Start with outcomes, not tools; prompts should map to your learning objectives and Bloom levels. Use chain-of-thought sparingly and never grade it; grade the final work against transparent criteria. Retrieval practice still wins—space it over days and mix in short, targeted hints from the model.
Teacher time is precious; automate the repeatable, keep judgment and pastoral care human. For accessibility, provide multi‑modal options: text, audio, and captioned video instructions.
Guard rails: forbid disallowed sources, cite where appropriate, and log versions for academic honesty. Honor privacy: minimize personal data, use district accounts, and rotate identifiers in exports. Document your playbooks; new colleagues and substitute teachers should onboard in one afternoon.
Clarity beats cleverness—if a student cannot restate the task, the prompt is too ornate. Always show a model answer and the rubric; feedback becomes legible and less surprising. Cold prompts underperform; prime with prior knowledge and short exemplars before free response.
Checklist
AI is not a shortcut to learning; it is a mirror that requires better questions and stronger rubrics. Start with outcomes, not tools; prompts should map to your learning objectives and Bloom levels. For accessibility, provide multi‑modal options: text, audio, and captioned video instructions.
Honor privacy: minimize personal data, use district accounts, and rotate identifiers in exports. Always show a model answer and the rubric; feedback becomes legible and less surprising. Clarity beats cleverness—if a student cannot restate the task, the prompt is too ornate.
Document your playbooks; new colleagues and substitute teachers should onboard in one afternoon. Retrieval practice still wins—space it over days and mix in short, targeted hints from the model.
- Define objectives; align prompts to verbs and outcomes.
- Provide exemplars; publish rubrics next to tasks.
- Decide what is allowed; teach citation and logging.
- Pilot with one class; iterate weekly based on evidence.
Conclusion
Clarity beats cleverness—if a student cannot restate the task, the prompt is too ornate. Guard rails: forbid disallowed sources, cite where appropriate, and log versions for academic honesty.
Cold prompts underperform; prime with prior knowledge and short exemplars before free response. Always show a model answer and the rubric; feedback becomes legible and less surprising. For accessibility, provide multi‑modal options: text, audio, and captioned video instructions.
Teacher time is precious; automate the repeatable, keep judgment and pastoral care human. Use chain-of-thought sparingly and never grade it; grade the final work against transparent criteria. Retrieval practice still wins—space it over days and mix in short, targeted hints from the model.