Typically teaching worldwide follows the principle that “If teachers teach, students learn. If students haven’t understood, teach them again.” This principle often fails. One of the reasons in our experience why this approach doesn’t work is because as part of the learning process, students develop misconceptions/alternate conceptions or harbour certain wrong notions which provide a hurdle in learning the concepts. While the development of these alternate conceptions cannot be prevented, they can certainly be handled using appropriate remedial techniques. But to do that, the first step is to know the exact error the student is making. A publicly available repository of student errors and misconceptions is what doesn’t currently exist. If compiled in a systematic manner using large-scale student responses, it can be an extremely powerful resource for teachers, educators, textbook writers and content developers who can then develop appropriate remediation strategies.
Taking inspiration from Professor Fei-Fei Li’s work on ImageNet, for a NeurIPS 2020 conference workshop, a competition was announced inviting ideas for ImageNet-like datasets that can serve as “benchmark” challenges in education. EI’s idea on ‘ImageNets for Math Errors’ got selected for one of the breakout sessions during the workshop. Here’s the idea we proposed – “the idea is to invite people around the world, whether it be teachers or parents, to share images and videos of student errors in basic arithmetic operations, which can be used to derive a database of common errors. Accuracy and fluency in arithmetic operations being key foundational numeracy skills, having a database of errors of this sort can be extremely valuable to teachers, textbook writers, content creators and researchers alike.
Here’s a clip of the breakout session from the workshop – https://slideslive.com/38942206
(Nishchal Shukla of Educational Initiatives in conversation with Samuel Ching of Schmidt Futures)