Chinese research outlines three layers of trustworthy AI
Summary and headline written by AI from the source article. How we work
Fei Song, Michael Dunn, and Zhenni Hu of the National University of Singapore and Nanjing Normal University analysed 43 Chinese-language articles and four policy reports to create a framework for trustworthy AI.
The researchers found Chinese scholars extensively study trustworthy AI but their work remains largely unexamined in English-language research. They organised these findings into a three-layer framework, arguing that trustworthy AI requires consideration of technology, governance, and human interaction.
The first layer focuses on technical aspects like reliability, transparency, and fairness, with Chinese researchers exploring applications in areas like personalised learning and financial security. The second layer concerns governance and regulation, reflecting China’s active policy discussions on AI, including balancing innovation with safety and determining liability for AI failures. The third layer addresses how trust forms during human-AI interaction, drawing from psychology and communication studies to understand how people calibrate their reliance on AI systems.
The team determined that an AI system must meet requirements across all three layers to be truly trustworthy. They also noted that Chinese scholarship offers unique perspectives on the philosophical debate about whether trust is appropriate when applied to AI, suggesting that AI can serve as a trustee. The researchers hope future work will integrate these perspectives and test the framework’s applicability across different cultures and regulatory environments.
