How to achieve ethical Artificial Intelligence? A systematic literature review on current academic frameworks and guidelines on AI ethics

The use of artificial intelligence (AI) technologies is increasing rapidly, affecting both personal life and the whole of society. Recently, the public discussion and research have paid more attention to ethical issues the use of AI can cause. Many frameworks and guidelines have been published on ethical AI during the past few years by governments, industry companies and non-governmental organisations. Previous research has focused on identifying ethical issues and reasons behind them and evaluating current frameworks and guidelines. This paper studies the research question: how to achieve ethical AI? This question is approached with a systematic literature review on current academic frameworks and guidelines related to the ethical AI. The focus is on the practical view the frameworks and guidelines provide for ethical design, development, use and governance of AI systems. As a result of the literature review, four themes could be identified. First, ethical values can be translated into design requirements with the help of value sensitive design and values hierarchy. Second, development of ethical AI is supported by multistakeholder approach and coordinating stakeholder impacts and expectations. Third, regulation and data governance are needed to prevent ethical risks of AI. Fourth, ethical dimension must be incorporated into management decisions related to the development and use of AI systems. These practical implications could be further concluded into a framework of different responsibilities for achieving ethical AI. The results of this research clarify responsibilities for regulators, designers and developers, organisations using AI and stakeholders and support previous research by addressing ethical values found before