AI 4 Education

By Pierre-Adrien Hanania and Farah Dabdoub.

As the guardian of societal progress and knowledge, education is one of the core CSD Indicators of Sustainable Development within the United Nation’s Agenda 21 (United Nations Department of Economic and Social Affairs, 2001). Its crucial role as a primary human right and driver of progress in health and well-being, social equity, trust and stability, as well as empowerment resonates in shaping the course of national economies and long-term economic growth (United Nations, 1992; United Nations, 2020).

The Sustainable Development Goal 4 “Quality Education” emphasizes the educational inequality and calls for school enrolment rates at every level and inclusive quality education for all, including vulnerable children, children with disabilities or other special needs, migrant and displaced children, indigenous children, and deprived children in rural areas (United Nations, 2015). In particular, the goal aims at achieving free, universal primary and secondary attainment, and ensuring equitable access to affordable and quality technical, vocational, and tertiary education. It also calls for removing gender and wealth disparities (UNESCO, 2021; UNDP, 2021).

Despite the effort in the field of education towards a sustainable future and the progress in the expansion of universal primary education by 2030, significant challenges persist in the completion rates for secondary and tertiary schooling. Especially poverty, armed battles, and further regional crisis in developing countries hinder the progress of education (UNDP, 2021).

Education in a digital world – on an enhanced educational journey

Considering the digital era we live in, the emergence of big data is transforming knowledge environments such as the education sector. The role that data and artificial intelligence (AI) can play to benefit students and teachers in their learning and teaching journey is game-changing. By applying advanced learning analytics and extracting data-driven insights into student behaviors, needs, and skills, children’s learning potential can be unlocked, and the operational efficiency of the educational institutions enhanced.

Around the globe, this new potential has been infused in several country strategies – ranging from Sweden’s degree programs in AI fields as part of the “National Approach for Artificial Intelligence” (Government Offices of Sweden, 2018) to China’s construction of an AI park in the context of its “New Generation Artificial Intelligence Development Plan” (European Commission, 2018; Reuters, 2018; Reuters, 2018) and lastly the Netherlands’ STAP-scheme, a 200 million euro investment to offer AI training opportunities and digital skills for citizens, mentioned in the report “Strategic action for artificial intelligence” (Government of the Netherlands, 2019).

In light of this trend, the virtual workshop organized by Capgemini in partnership with the UN’s agency ITU on AI’s role in achieving the Sustainable Development Goals (SDGs) provided a snapshot of the impact that AI and data can have on educational institutions, with a session dedicated to this field.

To the blackboard with AI – the technology’s promises and potential

AI in education can be used to enhance student learning outcomes, promote personalization, and improve access to education, particularly for girls and vulnerable children. Along with Capgemini’s four PublicGoesAI playgrounds, AI technology can be seen in the light of the four following ways to enhance teaching and learning solutions.

1.      Intelligently automating education-related administration processes

A first lever to activate societal progress is the mix between the intelligent use of data and automation. When applied on heavy, routine tasks, it can take a bit of the burden lasting on administrations and help address all citizen queries in less time.

The Department for Education (DfE) in the United Kingdom, responsible for children’s services and education from early years through higher education and apprenticeships, aimed at reducing the need for manual processing of all forms of digital correspondence and enhancing its reaction rate to incoming emails. In cooperation with Capgemini, the Department established and implemented a best-in-class robotic process automation (RPA) solution processing incoming emails for faster follow-up and more time for the most crucial cases, emails concerning children at risk.

The question of how to allocate more time to the most crucial aspect of education is indeed a key to progress where AI can help. For instance, administrative tasks such as grading and curriculum development take up a lot of time and reduce teachers’ time for in-depth interactions of high quality with their students. The automatic curriculum creation or the implementation of an essay-grading machine that matches humans 92% of the time (UNESCO, 2019) – through augmented intelligence assistance – can help teachers devote their time to the class and respond to students’ needs.

2.      Augmenting the interaction with children

With the advances in mobile technology, real-time communication with people through live chat interfaces has become a common possibility, augmenting interaction channels for all actors of the educational environment.

In the education sector, scalable AI-based educational chatbots can offer personalized advice and support children and youths, specifically students with disabilities and health impairments, in their learning process. The virtual tutors have the potential to provide students and teachers with analytics on their learning (Muslim et al., 2020) and offer interactive foreign language training, helping students to enhance their reading, speaking, and writing skills (Ruan et al., 2019).

3.      Detecting educational anomalies

In many countries, classrooms are full, and teachers lack the time to truly accompany each of the students with the appropriate resources. This can lead to individual failures with a long-term impact, such as drop-out. Machine learning techniques can be used to look into patterns to identify real-time anomalies.

One of the main challenges currently facing schools is student dropout rates that have been increasing around the world. With regard to Europe, a concern is that on average 10% of the students drop out before obtaining a higher academic degree. Incomplete higher education can hamper the Europe 2020 strategy’s goal of “having at least 40% of 30–34-year olds complete higher education, thus increasing the overall education level.” (European Commission, 2015) Reducing dropout and improving completion rates in higher education is one of the key policies for achieving this target.

To support schools, Capgemini Netherlands developed a predictive model that, by exploiting machine learning techniques, enables early identification of students who might drop out. Considering approximately 33,000 sample observations and approximately 1,800 dropout observations, and taking into account the class imbalance, the prediction accuracy was 91% for observations that the model has not been trained on. Thus, the AI-based prediction tool allows for early interventions and helps teachers to give affected students more attention to ensure that they do not compromise their education.

4.      Helping schools and teachers in decision-making processes

Going one step further than the detection of anomalies, AI can function as decision support for educational leaders, helping them to leverage data and insights to make more informed decisions. This can take the form of predicting occurrence with pattern recognition, for example, to detect the risk of drop-out before it even happens.

In France, decreasing results and low scores of French students over the last few years in the domains mathematics and reading, slightly above the OECD average, was concerning (OECD 2019). Capgemini France collaborated with the Ministry of National Education to enable adaptive learning solutions based on evaluation of the impact of learning methods: The mathematics application within the framework of the PISA study generates millions of logs that were not analyzed until 2016. Data science algorithms were used to observe the impact of teaching methods by analyzing the results and logs on the same exercise over several years. This enabled revealing learning paths and evaluating the impact of learning methods.

Another brick in the wall? Risks to be addressed in AI4Education

In their quests for educational sustainability with AI, societies will need to assess where the impact of the technology starts and where it shall stop. Indeed, concrete roadmaps are yet to be established in regards to the protection of children’s data, privacy, and digital rights and it will be of great importance to elaborate on how to bridge the global digital divide to facilitate equal access to the advantages of AI applications.

As a matter of fact, the introduction of AI is challenging the digital divide – the gap between those who have access to information communications technologies (ICT) and those who don’t – since robust digital infrastructure and data are the backbones of AI technologies. Considering the low internet use in Africa with only 28.2% (ITU, 2019), the lack of connectivity and electricity in the least developed countries (LDC), students, above all girls, in developing and the least developed countries will be left behind, whereas educational institutions and the students with access to digital infrastructure “will be the first to reap the benefits of these technologies” (ITU, 2018).

Another concern relates to the rights of children in the age of algorithms, including protection of children against discrimination due to race, ethnicity, gender, or economic status. To this day, for instance, there exist school segregation and concomitant educational inequity in the United States due to ethnically concentrated districts. Considering that academic achievement is positively connected with job performance, an algorithm trained using only data on the location of high schools could assign a high rank to fewer students of color and fail to identify high-performing students despite the attendance of a less-advantaged school (Darling-Hammond, 2018). In this case, racial bias in AI systems adversely affects the future of minority students attending academically weak schools.

A further pressing concern bears upon the ethics and transparency in the collection of data and, above all, the protection of children’s data and privacy. Access to education systems and concentrations of personal student and teacher information, including addresses and academic performances, could increase the risk of cybercrime and lead to a misuse of data. Building on AI and data will need to go hand in hand with a resilience strategy against cyberattacks.

The final challenge, as in other spaces where the human is key, is to establish a sustainable relationship between humans and technology. Teachers, as well as students, need to be prepared for AI-based applications in schooling to use the technology beneficially. From a development perspective, countries and regions with less strong technical capacities will face increased difficulties when adopting these intelligent solutions in education. Furthermore, trusted AI involves looping in all actors in an early phase of the project, with workshops and change management.

AI for Education – Towards enhanced learning outcomes, equity, and inclusion

Education has always been and remains a significant key to escaping poverty and hardship. Yet, approximately 260 million children and youths had no access to education in 2018 (UNESCO, 2018). In addition, most states were forced to temporarily close their schools, as a safety measure resulting from the ongoing COVID-19 pandemic. The transition to online learning in this regard has exposed significant gaps in school systems and has proven the necessity for technology-assisted learning. The above-mentioned solutions showed that AI has the potential to accelerate progress towards the accomplishment of Sustainable Development Goal 4, which targets education.

However, the benefits of AI must be available to all children and adults. Therefore, responsive national policies need to consider the following aspects when incorporating AI in their education system and embracing the potential of big data and learning analytics.

• Providing educational institutions and the students with access to digital infrastructure to ensure equity and inclusion.
• Regulating AI in terms of accountability and transparency to eliminate biases in algorithms and protect students against discrimination due to race, ethnicity, gender, or economic status.
• Applying strong protection of education data to prevent misuse of data; anonymizing the information, so that students cannot be connected to the data; using data encryption so that data cannot be interpreted by the analyst.
• Building AI capacity beyond basic ICT competencies to prepare teachers and other educational stakeholders in the era of AI.

AI technology can be used in four ways to enhance teaching and learning solutions. Want to know more? Check out our AI4Education point of view here.

Please reach out to the authors for more information.