evidoria

← Back to browse

Good practice Imported

AI-Enhanced Early-Warning System for School Dropout — Quest Alliance and UNICEF's Pilot in Uttar Pradesh, India

India · Gonda · See the India profile

Evidence: Observational / pre–post Top 17% 67/100 · Ask Evidence Copilot about this practice

UNICEF and Quest Alliance built an AI-enabled early-warning system tracking attendance and risk indicators to flag Uttar Pradesh students likely to drop out. Pilot schools saw attendance rise from 48.8% to 62.8%, prompting expansion from 100 to over 8,300 schools.

48.8%
Attendance in pilot schools (before)
62.8%
Attendance in pilot schools (after)
150
Original pilot schools (2022)
8,300+
Schools scaled to
~72%
Gujarat Vidya Samiksha Kendra dropout-risk model accuracy
AI-Enhanced Early-Warning System for School Dropout — Quest Alliance and UNICEF's Pilot in Uttar Pradesh, India

Details

Maturity
Scaling
Promoter
Quest Alliance & UNICEF India, with Uttar Pradesh Department of School Education (Samagra Shiksha Abhiyan)
Period
2022–present
Keywords
school administration, dropout prevention, learning analytics, government partnership

Context

Quest Alliance launched an Early Warning System (EWS) for school dropout prevention in 2022 under its Anandshala programme, piloted in 150 government schools across the Devipatan division of Uttar Pradesh (Gonda, Balrampur, Shravasti and Bahraich districts) in partnership with UNICEF India and the state's Department of School Education / Samagra Shiksha Abhiyan.

Objectives

Combine attendance and performance indicators to help teachers 'understand students' needs better and build better relationships' with children at risk of leaving school, and flag at-risk students earlier.

Activities

UNICEF's subsequent reporting frames the evolved system as AI-enabled predictive analytics, developed jointly by State Departments of Education in Gujarat and Uttar Pradesh, feeding teacher-facing dashboards. Teachers and education authorities, not the algorithm, lead the resulting interventions.

Results

Attendance in Uttar Pradesh's pilot schools rose from 48.8% to 62.8%, results UNICEF cites as the basis for scaling the early-warning approach from around 100 schools to more than 8,300, with testing now under way in seven further Indian states.

Conclusions

This is observational pre/post pilot data rather than a randomised or controlled evaluation. A related but administratively distinct system, Gujarat's Vidya Samiksha Kendra, uses a comparable predictive model reported at roughly 72% accuracy in identifying dropout risk, but the Uttar Pradesh pilot is run as its own state-level initiative rather than a direct replication.

Implementation
Indicative cost
High (€500k–€5M) — Piloted in 150 schools then scaled to 8,300+ schools across the state and expanding testing into seven further states.
Time to results
Medium (1–3 years) — Piloted from 2022; scaled within a few years; still expanding as of the source reporting.
Staffing & skills
Quest Alliance, UNICEF India, Uttar Pradesh Department of School Education (Samagra Shiksha Abhiyan), classroom teachers

Conditions for success

  • Teachers and education authorities, not the algorithm, lead the resulting interventions
  • Dashboards designed to help teachers build relationships with at-risk students, not replace them
  • Built on existing attendance/performance data collection rather than new infrastructure

Common failure modes

  • Original 2022 pilot documentation does not itself detail an AI/ML methodology, only later UNICEF reporting frames it as AI-enabled predictive analytics
  • No published randomised or controlled evaluation; attendance gains are a behavioural proxy, not a direct learning-outcome measure

Where it fits

Governance type
NGO-UN-state government partnership
Scale
state-level pilot scaling toward multi-state
Income level
lower-middle income

Commonly funded by

National / regional programmes

Indicative funding routes for practices of this type — always check each programme's current calls and eligibility rules.

Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.

Data sources

Where this practice's information was retrieved from, and when.

Attachments

Similar practices you may find useful