Where did that student risk model come from? Hidden inference in agentic AI-generated teacher-facing dashboards

Agentic Artificial Intelligence (AI) coding tools have the potential to democratise the development of teacher-facing learning analytics dashboards by enabling educators to generate complete applications from natural language specifications. In this study we evaluate Claude Code (Fable) and OpenAI Codex (GPT-5.5 High) across three dashboard specifications of increasing complexity using realistic educational datasets typically available to educators. Both AI systems successfully generated functional dashboards that accurately implemented the required visualisations and interactions. However, they also independently introduced additional analytical functionality, including student risk classifications and AI-generated insights, despite these features not being requested. These implicit inferences raise important concerns regarding transparency, bias and student data privacy, particularly where educators may be unaware of the analytical models or external AI services embedded within generated applications. The findings suggest that while agentic AI can significantly reduce the technical barriers for end user educators developing learning analytics tools, careful review of generated analytical logic and data flows is essential to support responsible and trustworthy use in educational settings.

Authors

Institutions

Publication Details

Journal
ASCILITE Publications
Published
2026-11-17
DOI
https://doi.org/10.65106/apubs.2026.3335
Primary Topic
Online Learning and Analytics
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Where did that student risk model come from? Hidden inference in agentic AI-generated teacher-facing dashboards

Aneesha Bakharia, Linda Corrin
3 links
ASCILITE Publications
Online Learning and Analytics
article

Where did that student risk model come from? Hidden inference in agentic AI-generated teacher-facing dashboards

Aneesha Bakharia, Linda Corrin
article en

Abstract

Agentic Artificial Intelligence (AI) coding tools have the potential to democratise the development of teacher-facing learning analytics dashboards by enabling educators to generate complete applications from natural language specifications. In this study we evaluate Claude Code (Fable) and OpenAI Codex (GPT-5.5 High) across three dashboard specifications of increasing complexity using realistic educational datasets typically available to educators. Both AI systems successfully generated functional dashboards that accurately implemented the required visualisations and interactions. However, they also independently introduced additional analytical functionality, including student risk classifications and AI-generated insights, despite these features not being requested. These implicit inferences raise important concerns regarding transparency, bias and student data privacy, particularly where educators may be unaware of the analytical models or external AI services embedded within generated applications. The findings suggest that while agentic AI can significantly reduce the technical barriers for end user educators developing learning analytics tools, careful review of generated analytical logic and data flows is essential to support responsible and trustworthy use in educational settings.

ASCILITE Publications
Deakin University (AU), The University of Queensland (AU)
Quality education
Online Learning and Analytics
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon