School-Based Tutoring Using Technology

1. Literacy and numeracy levels in the communities that we serve are low, and trail behind international benchmarks. These low achievement levels contribute to a misalignment with curricula; textbooks and curricula are targeted towards higher-performing pupils, leaving struggling pupils behind. And within an average classroom, achievement levels vary widely, especially in upper grades. 

2. In order to create a more equitable classroom, we propose a technology-enabled tutoring programme. This programme would recruit and train community-based tutors. These tutors would then provide pull-out tutoring services within the school setting, with a focus on low-performing pupils. Tutors would be supported with structured, digital tutoring guides. These guides would be informed by classroom data, including item-level exam data.

3. We propose to (A) create a replicable blue-print for a scalable tutoring programme and (B) provide rigorous evidence that school-based, technology-enabled tutoring is a cost-effective strategy to generate gains in literacy and numeracy at-scale.

Primary pupils in LMICs experience low learning levels, and classrooms have a wide range of ability levels. This within-class heterogeneity increases as pupils progress through successive grades. Efforts to promote personalised instruction are complicated by large class sizes. 

To illustrate heterogeneity, see Muradilharan et al. (2018), as well as internal analyses from one NewGlobe programme:

A body of evidence highlights the benefits of teaching to the right level. However, scaling these initiatives has been challenging, due to cost, bureaucracy, and issues with scalability. Combining tutoring with the use of digital, structured pedagogy is one promising way to address these challenges. Tutoring can have an outsized impact on achievement (Banerjee et al. (2017); Elbaunz et al. (2000); Cohen et al. (1982); Gersten and Baker (2000)).

Despite a large evidential base, we find gaps in the tutoring literature. First, it disproportionately focuses on western contexts, whereas tutoring is most relevant for LMICs with larger class sizes and a gap between curriculum standards and true learning levels (Glewwe, Kremer, and Moulin, 2009). Second, research has not yet explored the use of digital, structured pedagogy in tutoring; this approach could be particularly impactful for higher grades (where more sophisticated pedagogical content knowledge is required).

We propose to implement and evaluate an innovative, school-based, technology-enabled tutoring programme, with four core elements:

• 1-1 and Small-Group Tutoring: Tutors, recruited from the local community, provide daily, school-based 1-1 and small-group tutoring. Tutoring to be offered in a pull-out setting during normal school hours. Lower teacher-pupil ratios will meet the needs of individual learners and address heterogeneous ability levels within a classroom.

• Induction and in-service teacher professional development: Tutors receive induction training from NewGlobe teacher trainers at the outset of the programme. This training would introduce core instructional techniques, specifically the 'Big Four' teacher skills. School leaders and classroom teachers provide ongoing in-service training.

• Digital, Structured Tutoring Guides: Tutors equipped with a tablet to access structured, digital tutoring guides. Guides are developed by NewGlobe staff, with extensive experience designing learning guides for traditional classroom instruction. Tutors will use the structured tutoring guides to provide technology-facilitated, personalised learning. Individualised instruction will target foundational literacy and numeracy skills, as well as curriculum-aligned content areas.

• Data-Driven Decision-Making: Tutors supported with data from NewGlobe’s ‘Let’s Mark!’ app, which captures and curates item-level data from ongoing diagnostic, midterm, and endterm exams. Machine learning system uses item-level data at-scale to target low-performing learning areas.

NewGlobe serves communities across 5 countries: Kenya, Uganda, Nigeria, Liberia, and India. These communities include private community school pupils (Bridge Kenya; Bridge Uganda; Bridge Nigeria; Bridge India) and government school pupils (government partnerships in Nigeria and Liberia). In all settings, we provide or support early childhood education (age 3-5) and primary school.

Pupils around the world in LMICs lack access to a meaningful, quality education, with many pupils in school but not learning. The World Bank (2019) estimates that 53% of pupils in LMICs cannot read and understand a simple story by the end of primary school (with that number as high as 80% in poor countries). While NewGlobe has achieved dramatic gains in the countries in which we operate, we encounter similar challenges in our communities. In Liberia, for example, only a quarter of adult Liberian women who completed primary education (through primary 6) can read (Liberia Institute of Statistics and Geo-Information Services, 2014).

Learning poverty is pervasive, and has significant ripple effects in a pupil's life. If a pupil is unable to read a story, that also has significant implications for their ability to meaningfully participate in other syllabus-aligned lessons, such as science or social studies. These challenges will compound over time, as the gap between the pupil's literacy level and the rigour of material (including textbook reading) widens. The same holds true for mathematics - a pupil who cannot do basic addition and subtraction cannot learn more complex, such as finding the perimeter of a square. 

These pupils require urgent support. But this is an extraordinarily difficult job for classroom teachers. First, they are tasked with covering a rigorous national syllabus, and often have no time to return to more foundational skills like oral reading fluency or 2-digit subtraction. Second, they have large class sizes with a wide range of ability levels. As such, they often resort to teaching to the median pupil, at the expense of pupils most at risk of falling behind.

NewGlobe has sought to better understand this challenge, as well as the broader needs of communities within each individual programme, through a variety of channels. These include extensive qualitative work within schools (daily lesson observations; focus groups; semi-structured teacher/principal interviews), qualitative outreach outside of the classroom (parent phone surveys; parent focus groups), and monitoring and evaluation of teaching and learning (school-level and classroom-level data, including attendance, teacher performance data, and exam scores). 

These ongoing efforts to better understand our communities (and specifically day-to-day operations and outcomes at our schools) serve to inform our collective approach to school administration. More specifically, however, they help us to identify new and pressing learning needs. In this case, the push to develop an adaptive tutoring programme reflects needs and desires of teachers, who communicate challenges addressing the wide range of ability levels in their classrooms. It also reflects trends in learning outcomes, including low learning levels in literacy and numeracy (especially relative to international benchmarks) and within-class heterogeneity that widens with each successive year (see previous response for an illustration of heterogeneity in fluency and numeracy in one NewGlobe programme). Finally, this solution addresses the desire among parents to see their children supported through personalised instruction at the school level.

The communities that we serve will play a major role in the development, support, and oversight of this programme. This will include up-front co-production of the programme (both the design of tutoring guides and of the training sessions for tutors) with teachers, school leaders, and tutors themselves; ongoing consultation with teachers, school leaders, and tutors regarding the quality of tutoring and the fidelity of implementation; and solicitation of parent and pupil feedback up-front and throughout the course of the programme. This engagement with various stakeholders will strengthen the programme, but also ensure that the programme serves the needs of children and parents.

NewGlobe serves low-income families in urban and rural communities. While we support pupils and families through a holistic education system, pupils still trail behind international benchmarks. We must reimagine the classroom in order to provide more personalised instruction that meets the needs of all learners. In any classroom, there is a wide range of achievement levels; these differences have been exacerbated by COVID-19 closures. A school-based, technology-enabled tutoring programme ensures quality learning that meets the needs of diverse learners. Training and digital, structured teacher guides would support tutors to support pupils of all ability levels, even in low-connectivity settings.

This is a 'concept' informed by significant institutional expertise.

As an organisation, we have employed versions of a tutoring programme in the past. Specifically, we have conducted two randomised evaluations of phone-based tutoring in response to COVID-19 closures. We have also evaluated (and subsequently scaled) a peer-tutoring programme across our Kenya community school network.

Our solution will certainly be informed by those evaluations and implementation experiences; but the intervention is fundamentally different, in that it is (1) in-person (as opposed to remote, for the phone-based tutoring) and (2) led by community-based tutors (as opposed to teachers, for the phone-based tutoring; as opposed to pupils, for the peer-tutoring).

Our organisation also has significant experience with other elements of this proposed solution, including instructional design and teacher training. NewGlobe has been designing digital, structured teacher guides for nearly 10 years and continues to train thousands of teachers around the world each year.

Tutoring itself is not new or innovative. In fact, conventional tutoring has been found to be so successful that organisations have sought ways to harness the power of tutoring in scalable group settings (see Bloom's 2 Sigma Problem). But to date, the '2 Sigma Problem' has been a conundrum for providers and practitioners. Despite numerous attempts to replicate Bloom's findings (derived from mastery learning and 1-1 tutoring), we have not yet found a scalable alternative that delivers such profound results. But the need for dramatic improvements in learning has never been so relevant, especially for struggling learners in LMICs.

Our solution will offer a blueprint for scalable, effective tutoring in LMICs. The solution combines the 1-1 and small-group attention necessary to support learners with mastery learning taking place during normal classroom instruction. It will also offer structured tutoring guides, up-front and ongoing training, and a system for leveraging classroom data to inform tutoring. This unique combination of three foundational elements of tutoring will not only deliver rapid and outsized learning gains for our pupils; it will help other organisations and governments to implement a similar programme.

Our goal is to not only deliver better learning for pupils in our communities; we want to initiate a global movement that better supports the needs of learners. Because of our work at scale, any programme would immediately reach 500,000 learners. But beyond that, our relationships with governments will help us to build bridges, share learnings, and facilitate uptake of a successful programme.

1. NewGlobe currently serves roughly 500,000 primary school pupils across 2,400 schools in 5 countries. These include pupils attending private community schools (~45,000), and a far larger number of pupils attending free government schools (~455,000). 

2. In one year, we expect to be serving roughly 650,000 pupils, likely in at least one new country. In five years, we expect to be serving 3 million pupils, the vast majority of whom will be attending free government schools (likely in ~3-5 new countries).

3. If this solution is taken to scale across NewGlobe programmes, it would impact each and every one of these pupils. 

Importantly, when we talk about how this programme 'serves' an individual, we are not discussing one-time usage of an app, or sporadic engagement in a programme. We are talking about daily, targeted support for learners through the formal school setting. As a result, an individual's engagement with the solution will be far more significant, given the way in which the solution will be situated within the broader NewGlobe educational ecosystem.

This solution is still in the conceptual phase, so progress towards impact objectives is not currently being monitored. But NewGlobe already has robust channels to monitor learning outcomes and lesson delivery, which will be used to monitor the tutoring programme.

Learning Outcomes

• Termly exam performance: Bi-termly midterm and endterm exams, which assess curriculum competency domains.
• Fluency assessments: Bi-termly oral reading fluency assessments, which measure correct words per minute.
• Diagnostic exams: Literacy and numeracy exams, which are used to assess foundational skills and level future supplemental literacy/numeracy programming.

For each of these three assessments, teachers administer and mark exams, and enter the scores into their teacher tablet. This data is then accessible via NewGlobe's central server.

In addition, Let's Mark! will provide even more granular data. The Let's Mark! app will capture and upload item-level data, offering an even more precise assessment of learning outcomes (for example, reading comprehension outcomes on a broader literacy assessment).

Implementation Monitoring

Quantitative: Because tutors will use digital tutoring guides, lesson delivery data will be accessible via NewGlobe's central servers. This allows support staff to monitor performance precisely at each individual school. In addition, Schools Supervisors will provide follow-up trouble-shooting and support.

Qualitative:  NewGlobe also has a field staff dedicated to observing lessons. This same system will be used to observe tutoring guides and provide feedback on (1) guide quality, (2) the channel through which pupil performance data flows to tutors, and (3) logistical elements of the programming (group selection for tutoring, transitions, etc.).

This solution is embedded within NewGlobe. NewGlobe is a for-profit social enterprise that (1) operates private community schools and (2) serves as a technical partner to governments on public-sector education transformation programmes. Any tutoring programme would be administered within the school setting of one or more NewGlobe programmes. 

Our proposed solution team would include:

• ~100 full-time tutors (50 in private school context; 50 in public school context)
• 2 full-time programme managers (oversight, analysis)
• 1 full-time instructional design associate (creation of guides)
• 1 short-term talent acquisition contractor (recruitment of tutors)

NewGlobe is uniquely positioned to implement and monitor an at-scale tutoring programme.

• Instructional Design: NewGlobe has a demonstrated track record designing digital instructional materials for in-person lessons and also for tutoring programmes (phone-based and peer-led).
• Digital Content: NewGlobe has an existing content management and distribution system to equip teachers with digital lesson guides aligned to contextualised, daily learning objectives.
• Teacher Training: NewGlobe has trained tens of thousands of teachers in numerous contexts, including government schools and private schools. This includes induction and ongoing professional development.
• Schools Supervision: NewGlobe has an existing cluster management and supervisory system in which Supervisors oversee small clusters of 8 schools; this ensures that individual programmes and systems are visited at least once every two weeks, and that school leaders are supported to promote high-quality instruction at their school.

Our team is comprised of experienced NewGlobe staff with years of collective experience in these relevant domains. The vast majority of our Solution team leadership and support staff live and work in the communities they serve. They have experience in their country's education system, as well as experience working in a solutions-oriented education transformation organisation.

In addition, the tutors themselves will play a meaningful role in the design and iteration of the programme itself. Teachers and tutors will co-produce the solution through frequent and meaningful consultation. NewGlobe field staff will observe tutors facilitating learning sessions and solicit ongoing feedback in order to ensure that materials and training are maximally relevant and supportive.

Diversity and inclusivity are central to our organisational mission, and to the solution team in particular. Our team will be diverse and include a variety of perspectives. Solutions will be driven by local leadership, including female leadership, from our regional support offices in the countries we operate, as well as from the schools themselves. 

• School-based support: School leaders and teachers will be the true drivers of the programme. They will support with the identification of tutors, and will also provide ongoing professional development for tutors.
• Regional support: A team based in regional support offices will provide local project leadership, including recruitment, training, monitoring, and ongoing support.
• International support: A small team from London, UK and Boston, USA will provide targeted support, including on financing and programme design.

NewGlobe programmes operate at-scale at the state or national level. This includes rural, peri-urban, and urban schools; Christian and Muslim areas; and lower-income and middle-income areas. As a result, the leadership driving this programme (school staff and tutors themselves) will be diverse in terms of gender, language, religion, ethnicity, income, and location.

In addition, hiring for new roles (including tutors and support staff) will provide equal opportunity to all candidates. We will not discriminate against candidates in any way. We will seek to promote diversity through inclusive hiring practices in order to assemble a team that reflects the diverse and vibrant communities that we serve.

NewGlobe is one of the largest education providers in the world. It has significant experience designing, implementing, and evaluating innovative education programmes at-scale. But NewGlobe also operates within a budget, which constrains our ability to invest in some unproven, cost-intensive programming (even if it might carry significant benefits to pupils and communities). 

We are applying to Solve because we believe technology-enabled, school-based tutoring presents a major opportunity to deliver dramatic learning gains at scale in LMICs. Pupils in LMICs are increasingly attending primary school, but an unacceptable number of pupils are in school but not learning. We must reimagine the school space and transcend the traditional notions of classroom teachers and pupils. Tutors would provide the necessary personalised instruction and targeted support, which would benefit all pupils but in particular pupils at risk of falling behind national and international benchmarks for learning.

We require up-front investment in order to demonstrate proof of concept for tutoring. We believe that a successful at-scale implementation, combined with a randomised evaluation of programme effects, would serve as a powerful signal to our government partners, potential funders, and donors. A successful proof-of-concept and large effects on learning would lead to larger-scale investment in the programme. 

We also want to capitalise on the Solve network, which can facilitate connections with other organisations and funders. Technology-enabled tutoring is a complex undertaking in terms of programme design and implementation; in addition, future funding will require a thoughtful strategy. The Solve network can be integral in supporting both objectives.

Business Model: Our solution team is comprised of experienced educational professionals, with diverse experiences ranging from instructional design to training to monitoring and evaluation. But we would benefit from support from individuals or organisations with more business expertise. This could include support and mentoring on how to refine the programme's business case or strategic development. This might also include networking and connections with potential funders in order to fund or subsidise future iterations of or investment in the programme.

Financial: This tutoring programme requires moderate up-front and ongoing investment to cover the costs of tutor salaries and a small but flexible programme support staff. We are keen to receive support on how to secure additional rounds of funding and support for this programme. This could include securing investment from international funders, private donors, or government funding of the programme. This might also include support identifying potential grants and funding opportunities to which we can apply.

Product / Service Distribution: MIT Solve has an extensive network of practitioners and innovators. While we have a great deal of experience designing and implementing programmes within a NewGlobe context, we are also eager for outside support and consultation with experts and innovators. This could include MIT experts on tutoring and personalised instruction, or people within the MIT network with experience implementing similar programmes at-scale. In the future, we also hope that our experience can spark uptake in non-NewGlobe contexts. The MIT network would be integral in facilitating these connections and conversations.

As far as we are aware, school-based tutoring is a relatively novel approach in LMICs. The closest parallel was outlined by Banerjee et al. (2017). We would be keen to connect with any of the co-authors in this space, including Abhijit Banerjee, Shawn Cole, Esther Duflo, or Leigh Linden. Specifically, we would benefit from conversations about the nature of the intervention itself, lessons learned from implementation, and advice on the most effective ways to evaluate a comparable programme.

We would also be eager to connect with J-PAL or its affiliated professors, and specifically those with a background in targeted instruction or tutoring. J-PAL North America has provided helpful resources in this space. 

More broadly, MIT Solve likely has connections in comparable spaces related to tutoring and programme design. We would benefit from connections brokered by MIT Solve, as we continue to hone our approach and implementation plan.

Finally, we would be eager to engage with those in the MIT Solve network with experience in machine learning. This will be useful as we design a strategy to harness an enormous amount of item-level exam data collected via Let's Mark! and curate this data for endline users (tutors). This will also be useful as we explore opportunities to shape the programme materials themselves (digital tutoring guides) using real-time item-level data collected at scale.

While NewGlobe does not exclusively serve refugee communities in emergency and post-conflict settings, NewGlobe does serve refugees attending government schools in a variety of settings. This is particularly true in Liberia, where a large number of refugee pupils attend government schools that are a part of the Bridge LEAP programme.

Technology-enabled, school-based tutoring is a particularly relevant strategy for struggling and diverse learners. This will likely include a large number of refugee pupils, who are often learning in an unfamiliar language, transitioning into a separate national curriculum, emerging from a period of missed learning, and experiencing a wide range of social and emotional distress.

If the Andan Prize for Innovation in Refugee Inclusion were interested in advancing this solution, we would be open to basing one of the initial 50-school pilots in Liberia, where we serve the largest number of refugee communities.

We also believe that this type of tutoring programme would have profound implications for refugee education programming in non-NewGlobe contexts. Refugees require especially personalised instruction to address their unique learning requirements. Yet all too often, they are expected to transition into an unfamiliar education system with minimal support structures in place. 

If we can establish a blueprint for a scalable and cost-effective tutoring programme, this could be borrowed in emergency and post-conflict contexts. And following any receipt of funding, NewGlobe would be keen to collaborate with organisations that more explicitly serve refugee communities in order to provide training and capacity-building on how to establish an effective tutoring programme.

Our solution is closely aligned with The GM Prize's priority areas, specifically (1) smart classrooms and (2) a focus on STEM education.

Smart Classrooms: Pupils around the world, and in particular in LMICs, are frequently in school but not learning. This is the antithesis of a smart classroom. As a global education community, we have also been lured into thinking that a 'smart' classroom is a 'technologically advanced' classroom, complete with devices, computers, and computer-assisted technology. Technology undoubtedly has a role to play in effective classroom teaching. But ultimately, progress will come from people and the community. A school-based, technology-enabled tutoring programme is supported by technology (tutoring tablets and digital tutoring guides; exam data capture to inform levelling and targeting), but driven by people (tutors recruited from the local community). And most importantly, this programme will ensure that each and every pupil in a classroom receives personalised learning opportunities. 

STEM Education: Progress on STEM subjects is crucial in order to prepare pupils for a 21st century labour market. But progress in STEM requires foundational literacy and numeracy skills. Pupils cannot learn engineering if they struggle to do basic maths. Pupils cannot learn physics if they do not know how to read a textbook with comprehension. In short, literacy and numeracy skills are essential in order to make progress on more advanced subjects. With GM's support, we can invest in the foundational skills that will enable a generation of pupils to meaningfully engage in STEM subjects at all levels of education.

Our solution will benefit all pupils, but particularly female pupils and those female pupils whose education has been disrupted. Extracurricular tutoring is a powerful learning opportunity, but it is more often selected for urban pupils and for male pupils. School-based tutoring is an equaliser. It is a response to low overall learning levels, and also a means to support pupils at risk of falling behind or dropping out. 

At present, the divide between out-of-school boys and girls is minimal (roughly 2% in Africa) (UNESCO, 2019). But girls are still at greater risk of dropping out. Providing more personalised and targeted learning opportunities for girls will ensure that they can meaningfully participate in the school environment. In addition, personalised learning will support girls who are over-age for their grade-level (specifically those whose education has been disrupted by early marriage or conflict). Girls who are over-age for their grade-level are far less likely to be attending school relative to their male counterparts.

Tutors will support female pupils, and specifically those who are over-age for their grade-level. These female pupils might struggle to return to in-person learning, where they have fallen behind grade-level standards. There is already a gap between current ability levels and curriculum rigour for all pupils. Time out of school will only compound that gap. Tutoring, specifically for female pupils, will accelerate learning, facilitate an improved return-to-school, and ease the transition back to in-person learning (for those female pupils whose education has been disrupted).

AI and machine learning has thus far been underutilised in the design and implementation of educational programmes. NewGlobe has made great strides towards empowering school leaders and teachers with learning and performance data. At present, NewGlobe is at the precipice of extraordinary progress in data-informed design and instruction. NewGlobe's Let's Mark! app is being rolled out at the time of this proposal. Let's Mark! will capture item-level exam data from hundreds of thousands of pupils. This data will provide an extraordinary (and thus far unseen) level of insight into pupil performance at a state-wide or nation-wide scale. 

But more importantly, this data could be used to inform the design of an adaptive tutoring programme. This would include:

• machine-learning to inform the design and delivery of tutoring guides; item-level data will serve to pinpoint the exact learning domains in which pupils are struggling (for example, is a pupil struggling with comprehension vs. vocabulary), and also identify pupils who require additional remediation.
• provision of item-level data by support staff to tutors themselves, who can use this data to target instruction at the individual or small-group level

No other education organisation can compare to NewGlobe's data infrastructure. This presents an extraordinary opportunity to target and align instruction to pupil learning levels. In addition, this is a compelling case in which big data can be used to advance more than just marketing and advertising; it can ensure that pupils receive the right level of instruction, and the opportunity to thrive in classroom learning environments.

Please see above for an explanation to the first question ('Explain how you are qualified for this prize').

We would use The AI for Humanity Prize to advance our solution in two major ways:

• Adaptive Guide Design: We would want to invest funds into R&D in order to explore ways in which item-level exam data can serve to guide the design and delivery of tutoring guides. More specifically, we want to investigate (1) how to design a tutoring template with hundreds of 'segments' and (2) how to use machine-learning to determine which 'segment' should be provided based on overall classroom performance OR on an individual pupil's performance on recent exams. Imagine a world in which that pupil's performance determines that they require support on word problems assessing perimeter and area. A 'perimeter and area' segment is then automatically assigned to that tutor's tablet for the relevant tutoring session. In this way, big-data and machine-learning are used in order to provide adaptive guide design and distribution to meet the needs of individual learners.
• Data Curation: We would also invest funds into R&D in order to explore ways to curate item-level data for teachers and tutors. This data loop would access item-level data from the Let's Mark! app (which captures item-level data), curate the data on NewGlobe's central servers, and relay the curated data back to teachers and tutors via the teacher tablet. In this way, teachers are equipped with high-impact pupil performance data that serves to guide instruction and tutoring.

While our solution is not specifically related to cryptocurrencies, we do believe it is innovative, sustainable, and targeted towards the needs of local communities. 

Innovative: Thus far, tutoring has been a privilege of those families with the economic means to hire a private tutor. And in addition, the division between in-class teaching and extra-curricular tutoring has meant that the two operate in silos, ignoring the obvious synergies between the two exercises. In addition, tutoring in LMICs has been an analog exercise. We propose to support school-based tutoring using the same technology and design principles that have powered outsized learning gains driven by classroom teaching.

Sustainable: This tutoring solution will require up-front investment in order to demonstrate a proof-of-concept. But the results will speak for themselves - we can achieve dramatic gains in learning outcomes (specifically foundational literacy and numeracy). And this proof-of-concept will help us to make the case to our government partners that a pathway to closing the achievement gap is through manageable investments in school-based tutors, supported by NewGlobe tutoring guides and technology infrastructure.

Targeting the Needs of Local Communities: While we propose to implement technology-enabled tutoring at a state-wide or national level, the programme itself will be locally-owned. Tutoring will provide individualised, relevant, and targeted instruction. Tutors will be recruited from the local community, providing capacity-building and a pathway to future opportunities in teaching. In addition, tutors will be identified by local school leaders and community leaders. This will ensure that tutors are respected local figures.