Curriculum-Grounded Teaching Design Assistant

Workbook-grounded and author-clarified; proposed reconstruction. Course and unit teaching design draws on curriculum intent, course and unit learning outcomes, assessment, policy, accreditation and existing practice. This authoritative context is distributed across documents and can be made retrievable during educator design work.

Proposed and author-clarified. Educators may design from memory or immediate preference when authoritative curriculum material is difficult to retrieve. Generic GenAI can propose plausible assessments and activities that drift from course intent or unit outcomes. Long undifferentiated prompts may also conceal which source supports an answer.

Proposed and author-clarified. Construct a bounded, versioned curriculum corpus at course and unit scale and configure retrieval that returns relevant passages with source links. Add an educator-controlled meta-prompt defining the design task, source hierarchy, alignment method, output and abstention rules. Use the grounded assistant to propose and critique assessment, activities and resources while separating retrieved evidence from interpretation. Require educators to verify source support and approve all teaching designs.

• A bounded, versioned and searchable curriculum corpus with source hierarchy and provenance.
• Source-linked answers and alignment maps for course- and unit-scale teaching-design questions.
• Reusable educator meta-prompts for assessment, activity, resource and curriculum-review tasks.
• A review record distinguishing curriculum evidence, interpretation, generated proposal and educator decision.

Course and unit leaders authorise curriculum sources and interpret their relationships. Educators define design questions and approve outputs; students and relevant stakeholders inform actual needs; governance, accessibility and curriculum roles review where required. The system retrieves, grounds, maps and proposes, while people determine educational purpose and acceptable trade-offs.

Reconstructed risks

• A neat curriculum map gives false confidence in actual learning or delivery capacity.
• Generated structures reproduce generic disciplinary sequences without local evidence.
• Alignment work narrows education to measurable statements and overlooks emergence or inquiry.
• Accessibility, cultural context or student voice is added late rather than shaping the scaffold.
• Policy or accreditation text is misinterpreted.
• Superseded curriculum material is retrieved as if it were current.
• Fabricated or misplaced citations give a generic recommendation false authority.
• The corpus contains confidential or irrelevant student information without a justified purpose.

• Minimum viable implementation: place a small set of current course and unit documents in a controlled assistant, create a readable educator meta-prompt and require every response to quote or link its supporting curriculum passages.
• Robust implementation: use structured ingestion, source hierarchy, metadata-filtered hybrid retrieval, version and supersession controls, citation validation, representative test queries, retrieval evaluation and documented educator approval.
• Low-code implementation: connect an approved document store, retrieval service and GenAI interface with visible source panels, reusable task prompts and an alignment-report template.
• Local or privacy-preserving implementation: keep restricted curriculum, assessment and accreditation material in controlled institutional storage; avoid ingesting student data unless separately justified and governed.
• Speculative implementation: maintain course- and unit-aware assistants that suggest assessments, activities and resource scaffolds across the curriculum, while requiring source-linked evidence, abstention on gaps and educator approval before use.