AI Framework for Scalable Automated Continuous Formative Assessment
Arjun Rajasekar, Sakshi Mallenahalli, Inzela Mirza, Praveen Kumar Palaboyina, Sai Kumar Pola, Syed Falahuddin Quadri, Aravind Gondi, Ramesh Loganathan · 2024
There has been a great push towards evidence based learning across the globe. However, the level of progress made towards this goal has been greatly varied. One of the main bottle necks to this progress has been the increased administrative requirements for implementation of evidence based learning. Continuous formative assessments (CFA) a key metric for implementation of evidence based learning, is a work load intensive process in its structure which has led to poor or incorrect implementations of continuous formative assessments in many schools. We present a framework that uses natural language processing and computer vision based tools to perform automated continuous formative assessment in classrooms.The framework analyses the video and conversation streams that occur during a classroom session, to assess the engagement between teachers and students and generate insights into the student learning and behaviour. The framework builds upon existing models for automatic speech recognition, speaker diarization, facial recognition, body pose detection, and large language models, to promise a scalable automated assessment tool capable of providing standardised continuous formative assessments. The study covers both an online or digital classroom scenario as well as an offline or physical classroom scenario. We present the framework as it currently stands and present key improvements to be made before the framework is viable for field use.