What is SafeInsights?

SafeInsights is a national large-scale education research hub that enables researchers to analyze learning data from multiple education data organizations (DOs, e.g., edtech platforms, educational institutions) in a secure, privacy-preserving environment. Your research code runs within protected data enclaves, and you receive only approved, aggregated results—you never see raw student data.

Our mission is to advance research that drives better learning for everyone.

SafeInsights is housed at Rice University and funded by a $90 million investment from the U.S. National Science Foundation. The project is a multi-institutional collaboration including:

  • Leading edtech platforms, educational institutions, and other Data Organizations
  • Privacy and security experts
  • Learning scientists, education, and AI/ML for education researchers
  • Legal and ethics scholars
  • Community engagement leaders
  • Technology infrastructure specialists
  • and many more

How It's Different

SafeInsights inverts the traditional research model: instead of you receiving a data export, you send your research queries to the where the data is collected.

Traditional ApproachSafeInsights Approach
Data exported to youYour analyses code goes to the data
You see individual recordsYou never see raw data
Privacy risks if leakedData never leaves secure environment
Each edtech/Data Organization(DO) has different research request processStandardized process across edtech/DO
Limited sample sizesAccess to millions of learners

The Core Innovation

SafeInsights uses a novel approach: "We open the data while releasing none of it."

Instead of giving you data:

  1. You write analysis code (in R, Python, SQL) on the SafeInsights Integrated Development Environment (IDE)
  2. Submit it to the DO of your choice via the SafeInsights Research Portal
  3. Personnel at the DO run your code in the secure enclave
  4. Only approved, aggregated outputs are released to you

What You Can Do with SafeInsights (Version 1)

SafeInsights currently supports post-hoc or secondary analysis of existing data on SafeInsights member DOs. This means, depending on the DO choose to work with, you can:

Analyze Engagement Patterns

Study how learners interact with educational content over time

  • Time on task
  • Resource usage patterns
  • Navigation behaviors
  • Completion rates

Examine Performance Relationships

Investigate connections between learning behaviors and outcomes

  • Practice frequency and test scores
  • Resource use and achievement
  • Learning strategies and retention

Compare Across Groups

Explore differences in how different populations engage and succeed

  • First-generation vs. continuing-generation students
  • Different prior preparation levels
  • Various demographic groups (where data available)

Build Predictive Models

Create models to forecast outcomes from historical patterns

  • Early warning systems
  • Success prediction
  • At-risk identification

Track Longitudinal Trends

Follow patterns over time within courses or across terms

  • Engagement over semester
  • Learning curve analysis
  • Retention patterns

Learn more about post-hoc research patterns →

What's Not Available Yet

SafeInsights Version 1 focuses exclusively on post-hoc analysis. We do not currently support:

Interventions or A/B tests - You cannot modify the student experience or randomly assign conditions

Cross-platform studies - You cannot combine data from multiple education platforms in a single study (called "fusion")

Direct data access - You cannot download or directly view individual student records

These capabilities are planned for future releases, stay tuned!

Who Uses SafeInsights

SafeInsights is designed for:

Education Researchers studying learning at scale

  • Graduate students conducting dissertation research
  • Postdocs investigating learning patterns
  • Faculty researchers with questions about digital learning

Data Scientists exploring educational data

  • Methodologists testing new analytic approaches
  • Learning scientists building models
  • Quantitative researchers needing large samples

Edtech Researchers at partner organizations

  • Internal research teams at education platforms
  • Product researchers validating features
  • Data analysts supporting evidence-based design

Scale and Reach

SafeInsights provides access to data from:

  • Millions of learners across K-12 and higher education
  • Multiple subject areas including STEM, humanities, social sciences
  • Diverse contexts from homework systems to full course platforms
  • Longitudinal data spanning months to years

Explore the Data Catalog →

Built on Strong Principles

SafeInsights operates according to principles of:

Privacy First

  • Student data never leaves protected enclaves
  • Researchers never see individual-level information
  • Multiple layers of protection and review

Learn about privacy protections →

Research Quality

  • Rigorous proposal review process
  • IRB oversight required
  • Open science practices encouraged
  • Reproducibility supported

Typical Research Timeline

From initial idea to receiving results:

  • Initial research request review: X-X days
  • Code review: X-X days/weeks
  • Output review: X-X days/weeks

Total: Average ETA for full research cycle between X-X weeks

The timeline varies based on study complexity and DO review processes. Early feasibility checks help avoid delays later.

Common Questions

Q: Will I ever see individual student records?
A: No. You receive only aggregated results that meet privacy thresholds (typically minimum cell sizes of 10-20). Individual records never leave the secure enclave. You will have access to synthetic data samples to draft your analyses based on.

Q: What programming languages can I use?
A: At the moment, we are supporting Python, R, and SQL. Over time, we will be able to offer support for other analytical tools of choice. You write your analysis using simulated data, then your code runs on real data within the enclave.

Q: Does this cost money?
A: SafeInsights access is free. Some studies may incur computational costs charged by the data organization (at cost, not profit). Most small-to-medium studies have minimal or no costs.

Q: Do I need IRB approval?
A: Yes, in most cases. Even though you don't directly access identifiable data, you're still conducting human subjects research that requires institutional ethics review.

See full FAQ →

Next Steps

Understand the approach

Explore what's available

Design your study

Get support

  • FAQ - Quick answers to common questions
  • Get help - Contact options for specific questions
  • Glossary - Key terms defined

This page last updated: December 2025