倫理的AIの設計を支援するためのAI

（老いぼれ人工知能研究者のプログラミング日記）

2019年3月8日

`Organic and dynamic tool for use with knowledge base of AI ethics for promoting engineers’ practice of ethical AI design'

と題した論文を発表しました。

Kaira Sekiguchi and Koichi Hori: Organic and dynamic tool for use with knowledge base of AI ethics for promoting engineers’ practice of ethical AI design', AI & Society, Springer (2018). https://doi.org/10.1007/s00146-018-0867-z

「人工知能として認識されない人工知能の埋め込まれる社会に向けて」 で紹介した論文にも書いたのですが、今後、人工知能の引き起こす問題は人間だけで解決するには複雑になり過ぎるため、人工知能の引き起こす問題を解決するための人工知能も研究開発する必要が出てきそうです。
我々のこの論文は、その方向の第一歩を示すものです。

この論文の全文は、
https://doi.org/10.1007/s00146-018-0867-z
からダウンロードすることができます。

下に、アブストラクトを載せます。

In recent years, ethical questions related to the development of artificial intelligence (AI) are being increasingly discussed. However, there has not been enough corresponding increase in the research and development associated with AI technology that incorporates with ethical discussion. We therefore implemented an organic and dynamic tool for use with knowledge base of AI ethics for engineers to promote engineers’ practice of ethical AI design to realize further social values. Here, “organic” means that the tool deals with complex relationships among different AI ethics. “Dynamic” means that the tool dynamically adopts new issues and helps engineers think in their own contexts. Data in the knowledge base of the tool is standardized based on the ethical design theory that consists of an extension of the hierarchical representation of artifacts to understand ethical considerations from the perspective of engineering, and a description method to express the design ideas. In addition, we apply the dynamic knowledge management model called knowledge liquidization and crystallization. To discuss the effects, we introduce three cases: a case for the clarification of differences in the structures among AI ethics and design ideas, a case for the presentation of semantic distance among them, and a case for the recommendation of the scenario paths that allow engineers to seamlessly use AI ethics in their own contexts. We discuss the effectiveness of the tool. We also show the probability that engineers can reconstruct AI ethics as a more practical one with professional ethicists.