Educator-guided architecture separating artifact-grounded diagnosis, pedagogical workflow control, and student-facing LLM generation¶

Design · All designs
Evidence · 1 claim (1 mixed) · 1 study (1 design), q1 · 0 of 1 report an effect size · 1 claim rests on one study

Description¶

The architecture places a hidden, artifact-grounded diagnosis inside an explicit, stateful pedagogical workflow so the language model is not solely responsible for instructional strategy. As the article puts it, "The educator defines the stages, student actions, and disclosure limits, while the LLM selects or generates content within those constraints." The controller assembles the context available to the LLM from the ERD diagnosis, educator resources, and episode history, and logs stage, context, responses, and disclosed information for later review. This preserves student authorship while making the timing, specificity, and extent of automated assistance explicit and configurable.

Design Implications¶

Context¶

Requirements¶

Educator-authored requirements, rubrics, concept-aligned questions and resources, and workflow definitions that express governance over disclosure and progression

Constraints¶

Because the LLM may still produce content that exceeds stage limits, the system logs interactions for later review; staged disclosure cannot protect against wrong-relationship or wrong-concept selection, and the no-error route is unbuffered

Target Learners¶

Undergraduate and mixed-level database systems students

Learning Goals¶

Conceptual database design: translating requirements into entities, attributes, relationships, and semantic constraints in ER diagrams

Claims¶

Staged Disclosure Diagnostic Error Patterns [~W]

Examples¶

-

Key Sources¶

Sara Riazi and Pedram Rooshenas. (2026). An Educator-Guided LLM Pedagogical Agent for Scaffolded Feedback in Conceptual Database Design. https://arxiv.org/abs/2610.00870v1

Page metadata

Field	Value
`type`	`design`
`id`	`educator-guided-diagnosis-workflow-separation`
`title`	`Educator-guided architecture separating artifact-grounded diagnosis, pedagogical workflow control, and student-facing LLM generation`
`description`	`The architecture places a hidden, artifact-grounded diagnosis inside an explicit, stateful pedagogical workflow so the language model is not solely responsible for instructional strategy.`
`status`	`review`
`generated`	by `process:settle-candidates` on `2026-10-10`
`trust`	unverified — no one has confirmed this page's content
`sources`	1 — `sara-riazi-and-pedram-rooshenas-2026`

Frontmatter is the page's machine-readable record — see the Schema & Guide.