Explainable AI For Automatic Scoring Using Large Language Models

Automatic scoring has always been a point of interest in educational settings. With the recentadvancements in Large Language Model this work emphasize on automatic scoring strategies with Large Language models by providing the explainability to the automatic scoring task. This thesis explores different open-source large language models and their capabilities to perform automatic scoring. Since the automatic scoring is directly rated to students and their evaluation; the explainability, reliability and trustworthiness of automatically generated scores is crucial. This research is an attempt provide explainability to these LLM generated scores by aligning LLMs with human grading strategies with attempt to improve the automatic scoring accuracy. Finally, this thesis also presented the implementation of system "GPTest", an end-to end system to perform the automatic scoring using LLM and used in educational setting by teachers and students to streamline the scoring and assessment procedure.