AI Ethics: A Bibliometric Analysis, Critical Issues, and Key Gaps

AI ethics research has evolved through three distinct phases: from initial exploration (incubation), to focusing on human-like AI attributes, to the current emphasis on "human-centric AI systems" that prioritize human needs and values in design.

The "Collingridge dilemma" represents a fundamental challenge: early in AI development, we lack sufficient information to regulate effectively, but by the time we understand the impacts, the technology is too entrenched to control—highlighting the need for proactive rather than reactive governance.

Transparency and explainability remain critical unsolved problems: the paper identifies challenges in making AI decision-making processes understandable to users and regulators, which is essential for accountability and public trust in AI systems.

Justice, fairness, and privacy protection are interconnected concerns: the research identifies that protecting individual privacy while ensuring fair treatment across different groups requires balancing competing values, and current approaches have not adequately resolved these tensions.

Two major research gaps need urgent attention: the paper calls for more work on "large ethics models" (potentially AI systems trained to handle ethical reasoning) and on "AI identification" (distinguishing AI-generated from human-generated content), suggesting these emerging areas are underdeveloped despite their growing importance.