Skip to content

Intent-to-didactic-approach mapping structuring responses by Bloom level and typed chunk selection

Design · All designs
Evidence · no claims cited

Description

The system elicited the instructor's actual response practice into a small set of declarative didactic approaches (Definition, Identification, Illustration, Explanation, Comparison, Complex synthesis), each targeting a Bloom level and defining core chunk types that structure the generated response plus an ontology-enrichment strategy. Five pedagogical intents form the exhaustive set: DEFINE, EXPLAIN, ILLUSTRATE, COMPARE, and IDENTIFY, where "IDENTIFY is the inverse of DEFINE". Constrained retrieval selects content by metadata filters on chunk type and concept id rather than unconstrained similarity.

Design Implications

Context

Requirements

  • A knowledge graph of concepts with structural and didactic relations curated by the instructor
  • An editorial corpus segmented into nine typed chunks with concept-identifier metadata

Constraints

  • If constrained retrieval returns no chunk for a required type, the system returns a static reformulation message rather than silently switching to unconstrained retrieval

Target Learners

  • University students asking conceptual questions about course content

Learning Goals

  • Responses structured to the pedagogical intent and Bloom level of the student's question

Claims

Examples

-

Key Sources

  • Laurent Brisson, Maria Teresa Segarra and Gregory Smits. (2026). A didactical-driven teacher assistant for a dimensional modeling course. https://arxiv.org/abs/2607.22598
Page metadata
Field Value
type design
id intent-didactic-approach-mapping
title Intent-to-didactic-approach mapping structuring responses by Bloom level and typed chunk selection
description The system elicited the instructor's actual response practice into a small set of declarative didactic approaches (Definition, Identification, Illustration, Explanation, Comparison, Complex synthesis), each targeting...
status review
generated by process:settle-candidates on 2026-10-10
trust unverified — no one has confirmed this page's content
sources 1 — laurent-brisson-2026

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