DataCanvas-EDU: a four-phase agentic framework for instructor-guided synthetic data generation¶

Design · All designs
Evidence · 1 claim (1 for) · 1 study (1 design), q2 · 0 of 1 report an effect size · 1 claim rests on one study

Description¶

DataCanvas-EDU is an agentic framework in which the instructor specifies a teaching brief (business setting, learner level, permitted tools, assignment scope) and intended patterns through conversation, while an AI agent writes generation code, checks the exported data, and prepares reference analyses, assignments, and rubrics. As the paper states, "Four phases—Plan, Create, Verify / Test Analysis, and Evaluate—organize the process and support instructor review and revision." A shared case specification connects each intended pattern to its generation rule, comparison groups, measurement, and sample-size requirements, guiding checks and revision.

Design Implications¶

Context¶

Requirements¶

Instructor approval of the complete design before the agent begins routine programming and numerical checks; changes to a substantive pattern or its intended difficulty require the instructor's review

Constraints¶

The current workflow generates records from scratch and its numerical helper supports Python generation and a single CSV table; code execution is subject to the host environment's permissions

Target Learners¶

undergraduate business analytics students

Learning Goals¶

data investigation and visualization
formulating analytical questions and justifying findings

Claims¶

Agent Pattern Discovery Gap Windowdash Vs Olist [+M]

Designed-pattern authoring: linking each teaching pattern to its generation rule, check, and reference analysis

Examples¶

-

Key Sources¶

Bang An, Maria Hamdani, Joseph Fox. (2026). DataCanvas-EDU: An Agentic Framework for Instructor-Guided Synthetic Data Generation in Business Analytics Education. https://arxiv.org/abs/2609.19617

Page metadata

Field	Value
`type`	`design`
`id`	`datacanvas-edu-framework`
`title`	`DataCanvas-EDU: a four-phase agentic framework for instructor-guided synthetic data generation`
`description`	`DataCanvas-EDU is an agentic framework in which the instructor specifies a teaching brief (business setting, learner level, permitted tools, assignment scope) and intended patterns through conversation, while an AI ag...`
`status`	`review`
`generated`	by `process:settle-candidates` on `2026-10-10`
`trust`	unverified — no one has confirmed this page's content
`sources`	1 — `bang-an-2026`

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