ICTworks

As an education research community, we have achieved noteworthy success by increasing the number of children attending school around the world. The grand challenge for today’s development agenda is to improve learning for all, especially those at the base of the learning distribution. Leveraging dynamic and adaptive curricular practices in real time is one way we can support more and better learning where it is needed the most.

Our ability to surpass this learning challenge has been constrained by two principle elements:

• Timeliness of the data we collect, and
• Knowing how translate the data into action

To truly motivate learning, teachers need actionable data in real-time that allow them to adjust their classroom instruction according to their students’ needs.

A serious limitation is that there are not nearly enough trained professionals to physically create individualized learning plans for the millions of students (and out of school learners) around the world.

AI4ED: Artificial Intelligence for Education Development

One solution can be to incorporate adaptable learning models generated through artificial intelligence (AI) to inform curriculum and instruction decisions by generating more effective, personalized learning models through automated, machine learning algorithms.

One application of this approach already popular in US colleges and universities involves early alert systems to promote retention where teachers are prompted through easy to interpret data visualizations to assess student risk. The AI solution explores demographic and learning information data on the specific student, along with other students that exhibit similar learning styles, to derive precise actions the teacher (or student) can take to keep them on track.

It is the combination of recursive modeling through constant data capture with targeted feedback and guidance, in real-time, that sets AI-enhanced solutions apart from traditional learning interventions.

Another application of this approach serves to dramatically improve traditional assessments by scaling the education researcher. Good algorithms can make sense of massive amounts of data more efficiently than even a team of data analysts. The constant process of predictive modeling through machine learning rapidly converges to ideal learning solutions that iteratively build on themselves. As more data is collected the algorithm improves, enhancing the predictive power of the proposed learning model.

The iterative nature of machine-learning algorithms works as an alternative to the common approach of randomized controlled trials that require prohibitive resources in time, money and expertise, and generally are only applicable within the context in which they are deployed.

AI in support of the pro-equity learning agenda

While all students can benefit from AI-enhanced learning solutions, students at the lower tail of the learning distribution would be particularly supported through the emphasis on contextualized curricular modeling. Essentially, the system would simulate one-to-one tutoring for every student to support skills development where they need it the most.

Research on remedial tutoring has already produced impressive results for struggling learners in developing countries. To date, a big part of the conversation around equity in education has emphasized disaggregation of data. However, we are doing a disservice to the pro-equity agenda if we only use the more finely disaggregated data to compare how struggling learners are different from their higher-achieving peers.

Instead, AI-enhanced learning allows educators to develop better solutions that respond to the specific learner needs while accounting for diverse background characteristics. Further, improving support to struggling learners will serve to raise the floor of a country’s entire education system.

The bottom line for AI learning solutions

AI for education development is a transformative learning model that will accelerate progress in developing countries. The AI4ED approach works to converge curriculum, pedagogy, and assessment to create more personalized learning ecologies within low-resourced contexts.

By Nathan M. Castillo, Postdoctoral Research Fellow at the University of Pennsylvania Graduate School of Education with contribution from Antonio Quevedo of Academic Global.