A Systematic Review of AI-Driven Assessment Tools in Education.

Abstract

As educational institutions navigate the complexities of "Education 4.0," the traditional paradigms of assessment are undergoing a radical transformation driven by Artificial Intelligence (AI). This systematic review synthesizes findings from 155 empirical studies published between 2015 and 2025 to evaluate the efficacy, architecture, and ethical implications of AI-driven assessment tools. We categorize these tools into three primary domains: Automated Grading Systems (AGS), Intelligent Tutoring Systems (ITS) with embedded formative assessment, and Predictive Learning Analytics (PLA). The analysis reveals that AI-driven tools can reduce grading time by up to 60% while achieving inter-rater reliability coefficients of 0.90 or higher compared to human evaluators. Beyond efficiency, the integration of Natural Language Processing (NLP) and Machine Learning (ML) enables the delivery of "affective feedback"—real-time responses that adapt to a student's cognitive and emotional state. However, the study identifies critical "black box" challenges, including algorithmic bias against diverse demographics, data privacy vulnerabilities, and the potential for a "standardized student" bias. This article provides a comprehensive roadmap for educators and policymakers, advocating for a "human-in-the-loop" framework to ensure that AI serves as a catalyst for equity rather than a mechanism for digital exclusion.

Ähnliche Arbeiten