Trustworthy NLP Systems for Educational Decision Support: A Human Centered AI Approach

Abstract

Advances in Natural Language Processing (NLP) in education have led to the development of decision support systems capable of processing textual data such as student essay responses, learning feedback, academic records, and evaluation documents. However, most previous research has focused on improving model accuracy, while aspects of trust, transparency, fairness, and human involvement in decision validation have not been a primary focus. This research aims to develop a framework for Trustworthy NLP Systems for Educational Decision Support based on a Human-Centered Artificial Intelligence approach that positions teachers, students, and educational policymakers as key actors in the system’s design, interpretation, and evaluation processes. The novelty of this research lies in the integration of four key dimensions explainability, fairness, reliability, and human oversight into the NLP system architecture to support more ethical, transparent, and accountable educational decisions. The research methodology employs an experimental approach involving stages of educational text data collection, data preprocessing, text representation, NLP-based modeling, model performance evaluation, and system trust analysis through Explainable AI and fairness evaluation. The developed system is not only designed to generate educational classifications or recommendations but also to provide explanations for the model’s decision-making basis, thereby enabling human verification. The expected outcome is the creation of a conceptual and technical NLP model capable of improving the quality of educational decision-making without compromising ethical principles, accountability, and user-centricity. This research contributes to strengthening the direction of educational NLP development that is not only computationally intelligent but also trustworthy, inclusive, and human-centered.

GS Cited Analysis