Read and Lead Thursday
To discuss the book’s chapter on using AI to identify out-of-school girls, from Every Last Girl by Safeena Husain.
Key Takeaways
Inefficient Saturation Model: The initial door-to-door survey was unsustainable, taking 6 years per district and wasting resources on villages with no out-of-school girls.
AI-Powered Precision: A machine learning model trained on 10 years of data from ~1M households now predicts high-need villages, increasing the average number of girls found per village from 18 to 42.
Concentration of Need: A key insight revealed that ~40% of out-of-school girls live in just 5% of villages, making targeted intervention feasible.
“Combination Therapy”: The model’s success was amplified by aligning with government policies (RTE, Beti Bachao), which addressed systemic barriers such as school access and son preference.
Topics
The Problem: Inefficient Saturation Strategy
The initial strategy was a door-to-door survey in 13 districts, aiming for 100% saturation.
Inefficiencies:
Slow: Estimated 6 years per district (e.g., 100k homes).
Wasted Resources: ~20% of villages had no out-of-school girls.
Overwhelmed Teams: Some villages had 100+ girls for one volunteer.
The Goal: Find a faster, more cost-effective way to reach high-need villages.
The Solution: AI-Powered Prediction Model
The Idea: ID Insight proposed using a machine learning model to predict where out-of-school girls live.
The Data Goldmine: 10 years of survey data from ~1M households provided the necessary training data.
Predictor Variables: The model was trained on 313 indicators from the Indian Census and DISE, including:
Parent literacy, caste, income, and household size.
School proximity, infrastructure, and enrollment rates.
Human-Machine Synergy: The model guides volunteers to high-need villages, but human intelligence is still required for on-the-ground work and for navigating real-world challenges (e.g., mountainous terrain, difficult officials).
The Impact: Precision Targeting & “Combination Therapy”
Initial Test: The model found 50–100% more girls than the human-led survey in the same villages.
Improved Efficiency: The average number of girls found per village increased from 18 to 42.
Strategic Focus: The model enables targeting the 5% of villages that contain ~40% of all out-of-school girls, making the problem manageable.
“Combination Therapy”: The model’s success was accelerated by aligning with government policies that addressed systemic issues:
RTE (Right to Education): Mandated schools within 1km, removing distance as a barrier.
Rajasthan Govt. Scheme: Offered financial incentives for girls’ enrollment.
Beti Bachao, Beti Padhao: Tackled son preference and elevated the value of girls’ education.
Discussion: Human Cost & Ethical AI
Human Cost: The group discussed the emotional impact of reading about girls named “Falthu” (useless), reflecting deep-seated societal devaluation.
Ethical AI: The model was praised as a positive example of AI use, in which technology amplifies human impact rather than replacing it.
Next Steps
Brinda: Resolve the Zoom background issue with the IT team.
All: Meet next Thursday to continue reading the book.
