Access to online learning: Machine learning analysis from a social justice perspective.

Area of Science:

• Educational Technology
• Data Science
• Global Health

Background:

• Access to education is crucial for academic gains, but global inequalities limit online learning experiences.
• Low- and middle-income countries face infrastructure, content, and parental support challenges, exacerbated by COVID-19.
• Girls in these regions are disproportionately affected, highlighting gender-based disparities in educational access.

Purpose of the Study:

• To identify critical features influencing access to online learning using data mining.
• To understand the impact of global factors and individual characteristics on online education participation.
• To uncover insights beyond initial theoretical models for improving educational equity.

Main Methods:

• Analysis of a large dataset (54,842,787 initial data points, subsample n=5000) from an online learning platform.
• Development of a machine learning model, combining theory-led and data-led approaches.
• Utilizing Shapley values to determine feature importance for online learning access.

Main Results:

• Country differences, gender, and the COVID-19 pandemic were significant predictors of online learning access.
• Math ability, year of birth, session difficulty, month of birth, and time to complete sessions emerged as crucial factors.
• Data-driven insights revealed previously unexamined variables impacting educational access.

Conclusions:

• Global and individual factors significantly shape online learning accessibility.
• Addressing disparities in infrastructure, content, and individual abilities is vital for equitable education.
• Machine learning models can uncover nuanced factors critical for enhancing online learning opportunities worldwide.