“We will propose requirements for social decision-making by / with AI. ”

Social implementation of AI is progressing rapidly. AI will be responsible for everything from problem analysis to automated decision-making. Automated decision-making may be desirable for a commercial company. However, what should be automated decision-making for issues that affect society and some times on a global scale? How should we, non-AI specialists in humanities, society, science, medicine treat AI, including the problem of bias in the dataset already pointed out? What output do we want for AI decision support? 

This study investigates the problems and issues brought by automated social decision making, and by using thought experiments called “Hardest Choices”, we will propose requirements for automated social decision-making (support), including requirements of high-quality data sets and output for decision support. In other words, in this study, non-AI specialists will present the required specifications based on “materials that we can provide for AI” and “results required for AI”. This research will clarify the role of non-AI specialists in the AI era and frame the development direction of automated social decision-making.