Towards Concrete and Connected AI Risk Assessment (C$^2$AIRA): A Systematic Mapping Study

• DOI: 10.48550/arxiv.2301.11616
• Publication Date: 2023-01-01
• Authors: Boming Xia; Qinghua Lu; Harsha Perera; Liming Zhu; Zhenchang Xing; Yue Liu; Jon Whittle
• Keywords: Software Engineering (cs.SE); FOS: Computer and information sciences; Computer Science - Software Engineering; 0202 electrical engineering, electronic engineering, information engineering; 02 engineering and technology

The rapid development of artificial intelligence (AI) has led to increasing concerns about the capability AI systems make decisions and behave responsibly. Responsible (RAI) refers use that benefit humans, society, environment while minimising risk negative consequences. To ensure responsible AI, risks associated with systems' must be identified, assessed mitigated. Various assessment frameworks have been released recently by governments, organisations, companies. However, it can challenging for stakeholders a clear picture available determine most suitable ones specific context. Additionally, there is need identify areas require further research or new frameworks, as well updating maintaining existing ones. fill gap, we present mapping study 16 from industry, non-government organizations (NGOs). We key characteristics each framework analyse them in terms RAI principles, stakeholders, system lifecycle stages, geographical locations, targeted domains, methods. Our provides comprehensive analysis current state highlights convergence divergence among them. also deficiencies outlines essential concrete connected (C$^2$AIRA) framework. findings insights help relevant choose guide design future towards concreteness connectedness.