Explainable and responsible artificial intelligence in education: Ethical challenges in adaptive learning systems

Abstract

The fast development of adaptive artificial intelligence learning systems in education has been a cause of increased apprehension among the issues of algorithmic bias, loss of transparency, accountability, low level, data privacy, and the lack of human control. Although Artificial Intelligence in Education has enhanced individualized learning, learner modeling, predictive analytics, and intelligent tutoring systems, most educational technologies remain black-box systems that are difficult to ethically decide upon and trust. The literature review explores the way Explainable Artificial Intelligence, Responsible Artificial Intelligence, and Hybrid Intelligence can be used to curb these issues in adaptive learning systems. This review has a systematic review of the recent research on Explainable Artificial Intelligence, Human-AI Collaboration, Responsible Innovation, Educational Data Mining, Learning Analytics, and Trustworthy AI in educational settings, using the PRISMA framework. The review summarized the results of peer-reviewed articles about adaptive explainable AI, fairness mindful machine learning, transparency of algorithms, mitigating bias, data management, and human friendly AI in higher education, school education and digital learning platforms. The results show that Explainable Artificial Intelligence and Hybrid Intelligence models have the potential to enhance transparency, accountability, fairness and user trust in adaptive learning systems to a substantial degree. The methods identified as human-in-the-loop systems, neural-symbolic AI, interpretable machine learning, and student-centric AI models were identified to boost educational equity, teacher supervision, and responsible decision support systems.