How can education leaders leverage big data and AI when scaling?

Group Submission Type: Formal Panel Session

Proposal

Working in highly marginalized and data poor communities across four states of India, Educate Girls has been guided and enabled by the smart use of big data and machine learning, leveraged in partnership with IDinsight and Children’s Investment Fund Foundation. This decade-long collaboration has involved funding, evaluation, innovation in data science use and most importantly the expansion of a highly effective inclusive education program. To date, Educate Girls has mobilized over 1.8 million out-of-school girls for enrolment in government primary schools, seen high retention and attendance rates and supported over 2.2 million children with quality remedial learning opportunities.
The revolution in big data and artificial intelligence (AI) provides multiple opportunities for super-charging impact at scale. In our roundtable, we will draw on the Educate Girls program to present examples of how IDinsight uses big data and AI to help organizations, like Educate Girls, at scale to increase their reach, strengthen implementation efficiency, and improve cost-effectiveness. The panel will also go on to consider other emerging data science practices that have high potential for the education sector, with a particular focus on inclusive scale.
Monitoring and evaluation play a crucial role in helping social sector leaders to demonstrate ‘proof-of-concept’ to justify investment in scaling an intervention. As is evidenced by recent RCT results from joint work in Madhya Pradesh India, Educate Girls (education service provider), IDinsight (data science and evaluation team) and Children’s Investment Fund Foundation (donor) now have comparative evaluations of a pre-scale and at-scale version of the same program intervention. (See McManus, J, Shah, N.B. Sturla, K and Kitzmuller, L., 2018, Educate Girls Development Impact Bond Final Evaluation Report, accessed here 2024, and as yet unpublished, McManus, J., October 2024, The Audacious Project, Final Evaluation). However, there is limited precedent for how social sector organizations have leveraged data science during scale-up. That said we have promising emerging practice that will be shared through this round table discussion.

Educate Girls has been working in India for 16 years, an era that has seen great advances in the education sector and the country, from the Right to Education Act in 2009, to the widespread penetration of mobile phone technology and the emergence and vast adoption of social messaging platforms. It has also seen great challenges with and persistent social and economic barriers to girls’ access to education and of course the COVID 19 pandemic which reversed many of the hard-won gains of the previous decade.

Inclusive education is no longer a universal problem in a country like India, with high GER figures that mask the continued day to day reality of millions of children, especially girls, who in hot spots across the country are still denied their right to an education.

This makes the scale up of initiatives like Educate Girls all the more difficult, as we first have to know where to find the most vulnerable communities. Accurate targeting and efficient implementation are a prerequisite, if we are to affordably reach the most marginalized with the most effective and appropriate programs. Donors and implementers alike are faced with hard choices when working out how to enable rights at the scale of the need, balancing speed, inclusion and cost. To leave no girl, no child and no community behind and have a truly inclusive approach, in education and beyond, innovations in the use of data and technology are emerging as the fuel we need to supercharge our work.

Some of the approaches that will be discussed on the panel include:

Using machine learning for smart targeting: As organizations scale, they aim to reach as many eligible recipients of their program as possible. However, many organizations work with marginalized and vulnerable populations that are invisible in official government records. Machine learning can help organizations predict where these marginalized populations are found so that they can more effectively target service provision.

Satellite imagery: Smart targeting models, like those used by Educate Girls and others, typically rely on existing programmatic and administrative data to form predictions. However, in many environments this data does not exist, is out-of-date, or is of questionable quality. In such cases, we envision the potential for using satellite imagery to predict outcomes and pinpoint marginalized populations. Implementers use this information to more effectively target service provision.

Using AI-powered tools to strengthen implementation processes: While machine learning is increasingly used to determine how and where organizations scale, more recent innovations have created opportunities for organizations to use AI to strengthen their internal operations and implementation processes at scale.

Increasing internal access to data for decision-making: Organizations at scale generate large amounts of monitoring data, but often struggle with using that data to inform decision-making in real-time. We have found that even the best-designed dashboards often receive limited use. To help organizations better access and use internal data, tools can allow authorized users to query their M&E database in natural language, over WhatsApp.

Sub Unit

• General Pool

Chair

• Jeffery McManus, IDinsight

Individual Presentations

• Using machine learning to predict where to scale girls’ education programs - Akhil Agrawal, Educate Girls

• Emerging innovations for using big data and artificial intelligence to super-charge education programs - Ashwini Maslekar, Educate Girls

• Investing in data and evidence for enabling scale, a funder perspective - Stuti Tripathi, Children's Investment Fund Foundation