Learn Without Walls

Research & Publications

AI-Powered Cognitive Scaffolding Systems for Statistics Education

Research Overview

My dissertation investigates how AI-powered cognitive scaffolding systems can improve learning outcomes in introductory statistics, with particular attention to diverse student populations across different institutional contexts.

Core Question: Can AI provide adaptive, personalized support that helps students overcome common conceptual obstacles in statistics while maintaining trustworthiness, fairness, and educational effectiveness?

Why This Matters

Research Questions

Primary Research Questions

RQ1: Effectiveness

How does AI-powered cognitive scaffolding affect learning outcomes in introductory statistics compared to traditional online resources?

RQ2: Scaffolding Strategies

What types of cognitive scaffolding (minimal hints, moderate guidance, extensive support) are most effective for different statistical concepts and student populations?

RQ3: Institutional & Demographic Differences

How do diverse student populations across different institutional contexts (community college, research university, open enrollment) engage with and benefit from AI scaffolding?

RQ4: Trustworthy AI

What design features ensure AI tutoring systems are trustworthy, fair, accurate, and explainable across all student demographic groups?

The Platform as Research Instrument

I am building a free online statistics learning platform that serves dual purposes:

Educational Mission

Provide high-quality, accessible statistics education to anyone who wants to learn, regardless of financial means or institutional affiliation.

Research Instrument

Collect rigorous data on how AI-powered cognitive scaffolding affects learning, engagement, and equity in statistics education.

Ethical Research: All research participation is voluntary with informed consent. Students can use the platform whether or not they participate in research. Data is anonymized and protected according to UCLA IRB protocols.

What Makes This Novel

Cognitive Scaffolding, Not Just Q&A

Unlike typical AI chatbots that simply answer questions, this system provides adaptive scaffolding — progressive hints, Socratic questioning, and just-in-time support that fades as students gain mastery.

Multi-Institutional Comparison

By studying students across different contexts (community college, research university, open enrollment), I can understand how institutional factors and student demographics affect AI scaffolding effectiveness.

Trustworthy AI Framework

Working with Professor Guang Cheng's expertise in trustworthy AI, the platform includes monitoring for accuracy, fairness across demographic groups, explainability, and reliability.

Real-World Impact

This isn't just a lab study — real students learn real statistics while contributing to research that will improve education for future learners.

Current Research Projects

Active Research Program: Beyond the dissertation platform development, I am conducting multiple studies examining the impact of AI-assisted course design on student outcomes and instructional consistency across diverse institutional contexts.

AI-Assisted Course Design in College Algebra for STEM (MATH 4/4C)

Institution: Santa Monica College

Status: Data Collection in Progress (Spring 2026)

A comparative study examining the impact of AI-assisted instructor course design on student outcomes in College Algebra for STEM (MATH 4) with corequisite support (MATH 4C). The study compares Spring 2025 baseline data (no AI-assisted design; 38 students, ~50% pass rate) against Spring 2026 (AI-designed course materials; 32 students, data collection in progress).

Multi-Institutional Research Program

Institutions: Santa Monica College (SMC), Loyola Marymount University (LMU), Joyce University of Nursing and Health Sciences

Status: Various Stages of Data Collection

Additional studies examining AI-assisted instructor productivity, course material development, assessment design, and multi-institutional scalability across three distinct institutional contexts.

  • Community College (SMC): Open-access institution serving diverse student populations
  • Research University (LMU): Private university with selective admissions
  • Professional Graduate Program (Joyce): Nursing and health sciences graduate programs with working adult learners

Publications

Under Review

Income Inequality, Not Gun Policy or Mental Illness, is the Strongest State-Level Predictor of Mass Shootings in the United States: A Multi-Method Analysis (2018-2024)

Author: Safaa Dabagh

Journal: BMC Public Health (Springer Nature)

Status: Under Review

A multi-method analysis of 3,852 mass shooting incidents across all 50 U.S. states and the District of Columbia (2018-2024) using five complementary analytical approaches. The study finds that income inequality (Gini coefficient) is the strongest and most consistent state-level predictor of mass shooting frequency.

In Preparation

AI-Assisted Course Design and Outcome Consistency in Community College Statistics: A Longitudinal Analysis Across California's Placement Reform Era

Author: Safaa Dabagh

Status: In Preparation

A longitudinal analysis of 25 sections of introductory statistics (756 students) at a large California community college over eight academic years (2018-2026). The central finding: the AI-assisted phase produced a semester-to-semester pass rate range of just 8.3 percentage points — the lowest of any policy era.

Planned Publications (From Dissertation)

  • Effectiveness Study: "AI-Powered Cognitive Scaffolding in Introductory Statistics: A Randomized Controlled Trial"
  • Equity Analysis: "Does AI Scaffolding Reduce or Amplify Educational Inequalities? Evidence from Community College Statistics Students"
  • Trustworthy AI: "Designing Trustworthy AI Tutoring Systems: Fairness, Explainability, and Reliability in Educational Contexts"
  • Platform Design: "Open-Source Platform for Research on AI-Enhanced Statistics Education"

Theoretical Framework

This research draws on multiple theoretical traditions:

Methodology

Mixed Methods Design

Quantitative Component

  • Participants: 300-500 students across multiple institutional contexts
  • Design: Randomized controlled trial with multiple treatment conditions
  • Outcome Measures: Learning gains, concept mastery, engagement metrics, self-efficacy
  • Analysis: Multilevel modeling, causal inference techniques, equity analysis by demographic subgroups

Qualitative Component

  • Student interviews (n~30) across different demographic groups
  • AI conversation analysis using natural language processing
  • User experience surveys and open-ended feedback
  • Case studies of students with different learning trajectories

Timeline

Phase Timeline Key Milestones
Platform Development Fall 2025 Build learning platform, AI scaffolding system, data collection infrastructure
IRB Approval Fall 2025 Submit and receive UCLA IRB approval for human subjects research
Pilot Study Winter 2026 Test with 30-50 students, refine protocol, validate instruments
Main Study Spring-Fall 2026 Full data collection across multiple cohorts and institutions
Analysis & Writing Winter 2026-2027 Data analysis, dissertation writing, manuscript preparation
Defense Spring 2027 Dissertation defense and graduation

Researcher & Contact

Principal Investigator

Safaa Dabagh
UCLA Statistics MA | Seeking PhD Readmission
Mathematics Instructor, Santa Monica College
Email: dabagh_safaa@smc.edu

Academic Status: I hold an MA in Statistics from UCLA and am currently applying for readmission to the UCLA Statistics PhD program to complete my dissertation research. My active research program continues during this transition period.

Get Involved

For Researchers & Educators

Interested in collaboration, accessing pilot data, or learning more about the platform architecture? Get in touch — I welcome partnerships and knowledge sharing.

Back to Home