A Systematic Literature Review of Human-Centered, Ethical, and Responsible AI

Governance and fairness dominate current AI ethics research, but may leave blind spots: The review of 164 papers from major conferences reveals that research emphasis is heavily concentrated on governance and fairness, potentially neglecting other critical areas like privacy, security, and human flourishing that could have significant societal impacts.

Different research communities focus on different aspects of responsible AI: Specialized conferences like AIES concentrate on governance but rarely address privacy and security, while more established venues like CHI and CSCW maintain broader research portfolios, suggesting that responsible AI requires diverse perspectives across multiple research communities.

Current research may not adequately prepare for unforeseen AI consequences: The paper warns that the narrow focus on specific themes in responsible AI research may leave society unprepared for "potential unforeseen and unknown implications of AI," indicating a need for more exploratory and anticipatory research approaches.

Explainability remains underexplored despite its importance: While explainability is identified as one of three main emphasis areas alongside governance and fairness, the paper's recommendation to expand research on this topic suggests current efforts may be insufficient for helping people understand how AI systems make decisions.

Responsible AI research needs broader resource allocation and scope: The authors recommend that future work should diversify research focus to encompass privacy, security, human flourishing, and other underexplored dimensions, rather than continuing to concentrate resources on governance and fairness alone.