Student Work Images, Standards & AI-Generated Misconceptions

EdLight, PBC

⚡ AT A GLANCE

RESOURCE TYPE *

Dataset/Data Portal

POPULATION FOCUS *

Students Teachers

DATA AVAILABLE *

Quantitative Standards-Aligned

LAST UPDATED (DATE) *

2026 January 23

SUBJECT AREA(S) *

MATH

EDUCATIONAL LEVELS COVERED *

6-8

PLATFORM DESCRIPTION (BRIEF) *

The EdLight Research Portal is a secure, researcher-facing platform that provides access to up to 300,000 annotated student math work samples, metadata, and AI model-generated insights. Each image is tagged with misconceptions, content standards, and instructional context. The portal supports query filtering, comparison by subgroups, and visualization of learning patterns. Researchers can download de-identified data in analysis-ready formats to study historical trends and AI performance.

We support data extensions through custom annotation layers. Researchers can apply their own tagging schemas to existing work samples or request the ingestion of de-identified external performance data to enrich existing records for correlational studies.

DOWNLOADS AND LINKS

📄

Data dictionary

💡

SAMPLE RESEARCH QUESTION

How do teacher-generated tags and AI-predicted misconceptions align across different student populations?

1. Capabilities & Content

Resource Type *

Dataset/Data Portal

Data Fields (Demographic Options)

Age Gender Identity Biological Sex Race/Ethnicity ELL Status Disability Status

Linkable Identifiers Present? *

Yes

If 'Yes' or 'Partial', List Relationship Details

Students were given pseudonym identifiers which linked the data included in the final dataset.

Data Type / Format

Clickstream Data Assessment Responses Standards-Aligned

2. Research Potential

Example Research Questions

How do teacher-generated tags and AI-predicted misconceptions align across different student populations?

How do patterns of misconception vary across classrooms using the same curriculum?

What are the most common historical misconceptions in 6-8th grade math based on up to 300k+ samples?

Priority Research Areas *

School/System Improvement Assessment & Measurement Teacher Support & Orchestration Affect & Motivation Equity, Access & Inclusion

Construct(s) of Interest

Student Mathematical Thinking, Misconception Detection, Instructional Responsiveness, Equity of Access to Feedback, AI Interpretability and Trust, Data-Driven Teacher Decision-Making

3. Related Research Examples

EdLight Insights

other

This is a page regarding Using Student Work Data to Improve Teaching & Learning. Published in partnership with the Research Partnership for Professional Learning (RPPL).

4. Funding Opportunities

Open Grants

Open

AIMS EduData Spring 2026

Award Range: $10,000 - $400,000 Due: 2026-03-02

View Details

5. Access & Collaboration

Access Process

Expression of Interest IRB Required Data Use Agreement Consultation

Data Access Process / Timeline

1. Submission of research proposal and IRB approval.

2. Data use agreement execution.

3. Secure portal account creation.

4. Access to historical datasets for download and analysis

External researchers are supported through structured onboarding, Q&A sessions, and data office hours. All image data is de-identified prior to access. Use cases with minors include IRB and district approvals where required. Role-based access ensures only authorized researchers can view or export subsets.