AI-DRIVEN SUBJECTIVE ASSESSMENT IN ONLINE EDUCATION: CHALLENGES, SOLUTIONS, AND FUTURE DIRECTIONS

Abstract

The increasing reliance on online education has increased the need for efficient and fair assessment methods to assess students’ subjective responses such as essays and reflections. Manual scoring of these tasks remains time-consuming, inconsistent, and difficult to scale. Artificial Intelligence (AI), particularly through Natural Language Processing (NLP) and Large-Scale Language Models (LSM) such as GPT and BERT, offers a potential alternative. This systematic review synthesizes findings from 30 peer-reviewed research articles (2020– 2025) examining AI approaches to automated subjective assessment. Through thematic synthesis, three critical issues are identified: (1) transparency and explainability, (2) contextual understanding, and (3) lack of uniform implementation guidelines. The review identifies key challenges, synthesizes emerging solutions such as Explainable AI (XAI) and human–AI hybrid models, and outlines future research and policy directions for AI-driven subjective assessment. A conceptual framework is also proposed to map the relationship between AI technologies, pedagogical principles, and institutional trust. This study contributes by providing practical guidance to educators, researchers, and policymakers in designing AI-based assessment systems that are transparent, fair, and responsive to context.