India’s Rajasthan state recently created a new state-of-the-art in large-scale assessments. In the year 22-23, the government conducted three competency-based assessments involving nearly 5 million students, probably the biggest-ever sample size in educational assessments. The education department in Rajasthan state, India, assessed all 5 million grades 3-8 learners by using AI and OCR technology. By combining paper tests with computer vision AI, the state collected massive data without purchasing additional digital devices. Teachers captured photos of paper tests using their mobile phones, and AI scored the papers automatically. Every child was able to participate in the program because digital devices were not needed for them. This effort created a world record for the largest-ever use of AI in public education (read the news story).
If you want to know more about the program, do read blog posts by the Director of the School Education Department in the Rajasthan state Gaurav Agrawal: (1) राजस्थान के शिक्षा में बढ़ते कदम and (2) Biggest AI intervention in public edtech: RKSMBK. The blog posts share the motivation behind the program, detail how AI was used to collect the data, and how all stakeholders benefitted from the data.
This unprecedented initiative resulted in the creation of the most extensive longitudinal educational dataset ever, with over 600 million data points, comprising competency-based beginning, mid, and end-of-year assessments of 5 million students.
Data Collection Process
The data collection process of this massive initiative was quite simple, perhaps one of the reasons for its success. Students took their competency-based tests on paper, teachers scanned the papers through their mobile phones, and AI scored the answer copies immediately.
How large-scale assessment data was collected from paper from public school classrooms
The AI technology Smart Paper was used for the collection of the assessment data from paper. Nearly 40 million paper sheets were scanned by approximately 300,000 public school teachers through the government edtech platform RKSMBK. Smart Paper API was used to score the photos of the paper worksheets captured by the teachers.
From Bigger Data to Better Adaptivity
Data is the food of personalized learning systems. To build successful adaptive systems in education, we need to give the systems actionable data based on which technology can provide the needed digital support. By creating large-scale systems of educational data, we are potentially increasing the impact of digital education. Kudos to all of the risk-takers who created the new state-of-the-art in large-scale assessments.