Assessing Economic Essays with ChatGPT: Systematic Review and Preliminary Design

Abstract

This study explores the integration of ChatGPT in automated essay scoring (AES) through a systematic literature review (SLR) of 46 Scopus-indexed articles published between 2020 and 2025. The review identifies key trends in the use of large language models (LLMs), particularly GPT-3.5, GPT-4, and ChatGPT, for evaluating student essays across various domains. Findings show that GPT-based models deliver strong accuracy and reliability, especially when guided by structured rubrics and effective prompt designs. Despite these advantages, challenges remain in terms of scoring validity, model consistency, and generalizability across educational contexts. Based on the synthesis, this study proposes a preliminary rubric-guided AES model using ChatGPT. The model is designed to align essay scoring with pedagogical standards while offering scalable and transparent evaluation. This paper contributes by mapping current practices, identifying research gaps, and laying the groundwork for future development of standardized, domain-specific AES systems, particularly in economics education. Further empirical validation and field-based experimentation are recommended to ensure the model’s pedagogical soundness and practical effectiveness.