An application of natural language processing to characterize environmental policy making in Japan

The Paris Agreement in 2015 is the largest long term international agreement on global warming ever adopted. Global warming itself and stakeholders’ efforts to cope with it will have far reaching effects on all over the world in the long run. Because effects of environmental policies can be both positive and negative, environmental policy making requires careful discussion, taking various factors into account to realize desirable outcomes. In the environmental policy making process in Japan, governmental committees composed of experts such as representatives of influential groups and organizations are held regularly, and this is a process in which experts discuss the issues from various perspectives and incorporate their opinions into the policy. However, they are often criticized in that they fail to properly discuss the necessary perspectives. Is Japanese environmental policy making really characterized by narrow focus? If so, why does the environmental policy making process in Japan tend to be narrowly-focused? In this study, we proposed a method for processing and quantitatively representing discussion using a text-as-data method, and analyzed and evaluated discussions in Japanese governmental committees. Hand-coded data of all sentences in representative 27 meetings are used for machine learning to predict topics that speakers refer to and their discussion formats (oppositions, suggestions, etc.) in a wider set of 208 meeting related to environmental policy making. The output of the learning was used to quantitatively measure proceeding of discussion, and we identified key elements in discussion formats that are important in having multiple and long-term perspectives in environmental policy making in Japan. Furthermore, additional evaluation of similarity among sentences in the minutes, we attempted to visualize the whole discussion. We found that the fundamental problem with the committee discussion in Japan’s environmental policy making is a bias in focused perspectives rather than a lack of essential perspectives. Policy implications were drawn that agenda setting by the secretariat is an important factor for the discussion in the committees and that the way the members interact in the dialogue can also improve the discussion.