How AI Can Be Used to Improve Data Collection Systems and Surveys

• In Event: Using AI to Improve Foundational Literacy and Numeracy

Proposal

The advent of artificial intelligence (AI), particularly Large-Language Models (LLMs), has opened an expansive frontier in education systems, with potential implications on cost efficiency, pedagogical innovation, and systemic reform. As part of a panel discussion centered around AI's role in fostering foundational literacy and numeracy, we wanted to share our thinking and some of the work Laterite is doing in this space.

Laterite is a data, research and analytics firm, that works on research in the social sectors with a focus on East and West Africa. We have operations in Rwanda, Ethiopia, Uganda, Tanzania, Kenya and Sierra Leone. Education is one of our core areas of focus and one where we have had the chance to experiment with the potential of some of these new technologies.

In collaboration with Rising Academies in Sierra Leone, we are currently testing ways to streamline the student-assessment-to-teacher-feedback process using AI. We are working on a solution to integrate optical character recognition (OCR), optical marker recognition (OMR), automated scripts, and LLMs with paper-based surveys. The goal is to enable students in Rising Academy’s schools to take standardised paper assessments; using OCR/OMR these assessments are then translated into an electronic dataset; results are then processed using automated analysis scripts, aggregated organisationally, and distilled into school-specific feedback messages using LLMs. Feedback is then shared with teachers and head-teachers via WhatsApp. This cost-effective, AI-enabled method has the potential to facilitate large-scale assessments and can make data actionable for informed decision-making at every step.

As we work on AI use in improving data collection systems, we also identified potential area of work that is made possible with LLMs. LLMs open the door to a whole new class of data, making it possible to analyse large bodies of text that previously would have been intractable. There are many opportunities to leverage this in the education sector. One example is an activity we are doing in Rwanda aimed at identifying gender biases within the primary education curriculum (this is part of a larger research effort focused on gender in education in Rwanda). Using the capabilities of LLMs to sift through thousands of pages of text, we expect to be able to extract significant insights previously inaccessible due to the sheer volume and complexity of the information.

With the integration of LLMs, there also arise efficiency gains in the research process. At Laterite, we are leveraging LLMs to enhance research efficiency through the forthcoming launch of Laterite.ai. This service will provide researchers with access to LLM-based tools designed to expedite traditionally time-consuming tasks such as coding research instruments, bias checking, and analysing open-ended questions.

Despite the promising benefits, we underscore the importance of proceeding with caution. Known limitations of these models, data privacy issues, and potential biases in analysis require a mindful approach as we continue exploring the boundless opportunities AI presents in reshaping education research.

Author

• Dimitri Stoelinga, Laterite