Ethical considerations in AI-based user profiling for knowledge management: A critical review

Abstract

Artificial Intelligence (AI) enhances knowledge management systems by improving efficiency and personaliza tion, but its rapid adoption raises ethical concerns. This study examines the ethical considerations in AI-based user profiling for knowledge management systems, with a focus on academic environments. The review employed thematic analysis to summarize existing research on ethical challenges and proposed new ways to integrate ethical considerations into AI-driven knowledge management systems. The review analysed 102 peer- reviewed articles published between 2020 and 2024 from major scientific databases such as IEEE Xplore, ACM Digital Library, and Scopus. The findings show that privacy 27.9 % and algorithmic bias 25.6 % had major ethical concerns revealing disparities between theoretical frameworks and implementable solutions. Five key bias sources were also identified: data deficiencies, demographic homogeneity, spurious correlations, improper comparators, and cognitive biases. While 73 % of the reviewed frameworks acknowledge at least one ethical consideration, only 28 % propose practical strategies to address them. Some promising approaches include explainable AI techniques, privacy-preserving algorithms, and fairness-aware machine learning. However, there are still gaps in addressing the long-term societal impacts. The study recommends the implementation of an Ethical AI Feedback Loop (EAFL) system, which continuously monitors, evaluates, and adjusts user profiling algorithms based on predefined ethical metrics. Additionally, the study introduces the concept of "Ethical Debt" to quantify and manage the long-term ethical implications. These innovative approaches aim to integrate ethical considerations directly into AI-based knowledge management systems, promoting responsible and adaptable user profiling practices.