Beta
MIT

Pitch

Fluent in communication and collaboration, we put a tutor in the hands of every child, helping them meet goals and overcome challenges.

Description

Summary

Private tutoring has proven to be the most effective way to improve learning outcomes and does not require brick-and-mortar institutions unavailable to refugee children who are dispersed and often undocumented.  Erudite makes private tutoring accessible to every learner by augmenting teaching capacity through the power of artificial intelligence and messaging bot (chatbot) technologies.  Deploying a peer-to-peer tutoring AI platform in a plurality of subjects - from literacy and math, to social-emotional regulation practices and cultural assimilation - will deliver on our 3 Pillars of Refugee Education:

1. Inclusivity for all Learners – Educate all students in refugee-impacted areas, connecting them with tutors both within their local communities and outside, thereby promoting learning interactions across boundaries of language, gender, faith, and nationality.  

2.On-Demand & Relevancy – Apply intelligent and mobile learning opportunities to the accommodate refugee children transiency by giving on-demand access to education that is relevant to their realities and sourced from peers who are building their own capacity as peer educators.

3. Capacity Building – The scalability of a mobile, AI-based tutoring platform enables local refugee communities, which are fragmented in needs even across each tent encampment, builds capacity for knowledge access from each other. Providing peer support through matching, as well as an AI conversational platform that can communicate via SMS in a plurality of languages, expands gives refugee youth reach to capacity within global knowledge economies.

What are the key outcomes and impact of your solution?

Coupling mobile delivery with our AI-fueled 1-1 tutoring system is critical as it ensures that every refugee child, as well as adults who can be developed into future educators, has access.  As evidence to support this approach, studies estimate that at least 86% of Syrian youth living in refugee camps have access to mobile phones, and agencies have given them SIM cards as they cross borders.  Whatsapp and Facebook have been critical mobile platforms for refugees to learn from their peers how to gain access to safe passage - now, Erudite is positioned to deliver the mobile platform for refugees to learn from their peers how to promote mastery of topics critical for 21st century academic and career opportunities.  While reaching toward mastery of these core topics, the peer-tutors engaging on the Erudite platform will be diverse, including both genders and individuals across faith and nationality.  This direct technology-based implementation of our Inclusivity of All Learners pillar fosters dialogue critical for peace-building and narrative-based learning as diverse populations contribute to mass delivery of accessible 1-1 tutoring.  

 

Evidence also supports that the ultimate form of mastery of a given topic is when we teach it to others. Thus, the Erudite AI system is peer-driven, which enables us to deliver on our On-Demand and Relevant Learning pillar by building the tutoring capacity of any child to levels comparable to professional tutors. It is the peers of refugee children who will know how to integrate learning with personally relevant and socio-emotional understanding.  Additionally, it is these same peers that can reduce barriers of accessibility to tutors by becoming effective peer evaluators and facilitators of learning.  As a result, the Erudite AI system will empower refugee children as “Makers of their own Learning” within and across local communities of internationally displaced youth, as well as help refugee children catch up to proficiency required by formal school admissions in their host countries.


Erudite’s system delivers on our Capacity Building pillar of refugee education by ensuring the scalability of platform that aims to tutor using peer mastery matching features and via SMS in a plurality of languages.  The fragmentation of knowledge capital within each refugee community results in a missed opportunity for peers to leverage the knowledge base of their peers.  Meanwhile, differentiating for varying proficiency levels of refugee children, most of whom have not been in traditional school systems for several years, presents the need for connecting this existing knowledge capital to the individual needs of learners.  Erudite can not only build local capacity of each refugee community by matching peers based on the knowledge they can give and receive, but also scale this ecosystem by augmenting tutoring capacity from global populations of tutors through peer matching and delivery of SMS-based language plurality systems.  

What actions do you propose to realize your stated goals?

By using artificial intelligence to augment the capacity of students, we can lower the cost of private tutoring that would serve as a critical education for refugee children. In addition, pedagogically, we can reinforce the abilities of these peers or teacher-students by teaching others. Our system makes the whole process efficient and very easy to use over a simple chat interface that the vast majority of students already use in their daily habits when communicating with peers. No other technologies have more market penetration than smartphones, enabling us to reach more users throughout the world.

 

Private tutoring is the most personalized form of learning, and once we reach a critical mass of users, we will be able to cluster different groups of students’ learning styles and match them with the appropriate tutoring methodology and generate optimum feedback for their learning behaviors.

 

In reality, students have been helping each other for centuries, highlighted during the Golden Era of Learning with Greek philosophers and teachers at the School of Athens with Plato and Socrates. Only the affluent and nobles, like Alexander, could afford a private tutor like Aristotle.  Now, Erudite aims to deliver what used to be a luxury to refugee children who would most benefit from 1-1 tutoring that can help poise them as citizen learners, make up for lost years out of school, and enter formal education institutions to obtain credentials required for future academic and career opportunities.  Erudite has made the process accessible, simple, and efficient,  as collaboration among students is facilitated by the best practices of artificial intelligence in natural language processing and reinforcement and transfer learning.

 

Our AI technology analyzes the learning and knowledge models of students. When our AI discovers a mistake, it automatically matches the students, based on a plurality of data sets, with peers, classmates or tutors in real-time. Our AI system finds the mistakes in real time and connects the students to the available assistance they need; moreover, it also suggests solutions and responses for the tutor to respond with (which can be edited or ignored altogether) to make the process faster, smarter, stronger. During the same time, the student tutor is reinforcing his/her own knowledge and gains skills on how to properly explain, how to correct mistakes, and how to find the right solution.

 

Refugee children using Erudite will benefit from exponential gain as our

system get used by millions of users all over the world.  We will be able to use data science techniques and clustering algorithm to identify specific learning style amongst the student population, at which point we will be able to generate potential feedback that works better with different style of learners - this differentiation is critical for refugee children who, despite their given age, would be placed in lower grade levels in schools or even barred entry to school until they are able to catch up. No professional tutor on the planet would be able to have a large enough sample of the population to be able to adapt to this level of personalization that refugee children require. On the other hand, machines somewhat lack the feel of human, and as such we believe our system could eventually outperform both fully automated tutoring system such as those created by Mcgraw-Hill and Carnegie Learning as well as the human tutor taken separately. The end result of our technology could be a massive decrease of the cost of education for refugee children, which will benefit productivity and economies.

 

We will build an entire data pipeline to process the information.

 

Specifically, a student will use our Pupil application to submit a problem they are struggling with.

 

Module 1: QueryClassifier

Technology: TensorFlow and Python 3.0

 

Module 2: Matching Algorithm

Technology: Python 3.0

We are building this module in collaboration with the Institut of Data Valorisation (IVADO) under the supervision of Professor Andrea Lodi at Polytechnique Montreal)

 

Module 3:Mathematics Extractor

Technology: Python 3.0

 

Module 4: Symbolic Computation of Errors

Technology: Haskell (may be converted to Python or Scala)

 

Module 5:Generative Dialogue System

Technology: TensorFlow


 

Module 6: Knowledge Profiler

Technology: Python 3.0

 

Module 7:Pupil

Technology: Node.js, Matrix.org (Synapse)

This module is our chat system for peer-to-peer communication. The core is an open source project called Matrix.org (Synapse is the server technology itself). We have considered using existing chat system (slack, Facebook Messenger), but we realized, it would cause problems in the future. It is important for us to be able to integrate a decent mathematical editor in the messaging system, as well as keep the data private and proprietary.

 

We are using best practice of micro-server architecture, Software as a Service (SaaS). This way we can eventually license each API’s and module separately.

Who will take these actions?

In order to enter, distribute, scale, and tailor our technology to the user data, we will require access to refugee communities that are usually barred to unapproved visitors.  For example, entering the Zaatari refugee camp in Jordan, let alone distributing to them, requires governmental permissions.  As such, Erudite will require access to a network of decision makers and government level approvers to ensure that we have access to the refugee children as our users.  Additionally, Erudite will require introduction to and facilitation of commitment from local leaders that have assumed the role of speaking on behalf of the refugee communities we aim to support.  For example, many villages in Lebanon and Jordan where Syrian refugees have taken refuge have come under protection of the village leaders who ensure their safety, access to resources, and transportation to schools.  In relation to Erudite’s needs, these local leaders will be able to evangelize and encourage sustained and expanded usage of our mobile, AI-based tutoring system.

Target geography

Last year, the UNHCR reported that 80% percent of the world's refugees are hosted in developing countries, which the agency described as a "deep imbalance in international support."  While Erudite aims to target this gap by matching peers from refugee communities with peers from the rest of the globe, we want to first deliver of our mobile AI tutoring solution where we have local experience.  Our Chief Technology Officer formerly created a platform used by NGO’s to deliver cross-cultural training in Pakistan, which currently hosts 1.5M refugees.  Additionally, our Chief Strategy Officer, a 15 year veteran in the education space, directly delivers education programs to refugee populations in Jordan, as well as Turkey and Lebanon, which host 2M and 1.15M refugees respectively.  She is also from Iran, the world’s 4th largest host country at 1M refugees.  This local expertise poises us to understand how to quickly put Erudite in the hands of refugee children so that they may access 1-1 tutoring required to maximize informal learning hours with high-quality content, as well as catch them up to proficiency levels required to enter formal schools within their host countries.  These target geographies present regions of greatest needs for our high-impact, scalable, and mobile AI solution, as well as aligns with the expertise of our team that understands the cultural and government-driven intricacies of each of these host countries. 

What do you expect are the costs associated with piloting and implementing the solution, and what is your business model?

Currently, we have $1.4M committed to the development of our technology stack. We will require an additional $470K to customize our technology toward the refugees’ children use cases specifically.


This $470K will be segmented into

Distribution Costs = $100,000 over the next 4 years.

Customization Cost of the Technology = $50,000 (our API tech will be open source)

Server Maintenance & Operations = $100,000 for 4 years.

General Administration = $100,000 for 4 years.

Localization = $120,000 for 4 years.

 

Timeline

 

Phase 1 (2017) will launch Pupil and data collection while Phase 2 (2018) will optimize technology, market penetration, and data cleaning to maximize AI performance. Phase 3 and 4 (2019-2020), will conduct data analytics and customer segmentation.  To summarize:

1) Technical demonstration of the Minimum Viable Product with Generated Natural Language Scorer and dataset collection: Beta Phase with 10 refugee camps in 2017

2) Learning outcome assessment with a 50,000 student sample by the end of 2017

3) 250,000 student milestone with a 50% retention rate by 2018, optimization of the technology to achieve 25% valid feedback options

4) 800,000 student milestone with a 60% retention rate by 2019 and optimization of the technology to achieve 50% valid feedback options

5) 1.5M student milestone with a 70% retention rate by 2020 and optimization to obtain 80% valid feedback options.

1. Phase 1 (2017)

a. Data collection - build a working text corpus of user generated feedback.

Build a base of 1000 student to generate text corpus large enough to start doing an NLP and NLG.

b. Number of users required to build the database:

Milestone: 200,000 students 10-20 sessions/year/ 300 words per session/ ave. = 1.2 billion words base

2. Phase 2 - 2018-

a. Tech optimization

Related solutions

References

Vote234 votes
Question?
Thumbs up 15 supporters Support Solution
Solution Summary
Erudite AI: A Deep Learning, Knowledge Exchange Platform
Team Solution: Only members listed on the Solution's Contributors tab will be able to edit this Solution. Members can request to join the Solution team on the Contributors tab. The Solution owner can open this Solution for anyone to edit using the Admin tab.  
By:  Erudite
Challenge: Learn: Refugee Education
How can we improve learning outcomes for refugee and displaced young people under 24?