금융, 이커머스, 음악 서비스에서 설명가능 인공지능의 추천 설명 방식에 대한 방향 제안 : 위험감수 의사결정 횟수에 따라 구분된 추천 서비스를 기반으로

Background With the development of AI, problems such as black boxes continue. Therefore, the purpose of this study is to confirm that the explanatory method of AI is different for each field divided according to the degree of risk-taking decision-making. Methods In this study, experienced user in the fields of finance, e-commerce, and music were selected as interviewees. The survey and in-depth interview were conducted a face-to-face interviews, and the analysis of results was conducted using Glaser's grounded theory, a qualitative analysis method. Results In finance, what if explanation was preferred. It proposed a recommended explanation with detailed explanation based on personal information and various comparison functions. In E-commerce, why explanation was preferred. It proposed a quick recommendation explanation based on accurate data. In music, explanation based on society was preferred. It proposed a convenient, fast, or detailed information-based recommendations. Conclusion This study classified recommended services according to risk-taking decisions, confirmed that the preferred explanation was different for each field, and suggested the direction of the explanation method.