Probabilistic approach using Bayesian networks and its application to behavior change support technology

• Published in: Journal of the National Institute of Public Health Vol. 73; no. 4; pp. 283 - 291
• Main Author: MOTOMURA, Yoichi
• Format: Journal Article
• Language: Japanese
• Published: National Institute of Public Health 31.10.2024; 国立保健医療科学院
• Subjects: Bayesian networks; behavior change support; consumer / resident digital twin; probabilistic approach; ベイジアンネットワーク; 生活者デジタルツイン; 確率的アプローチ; 行動変容支援
• ISSN: 1347-6459; 2432-0722
• DOI: 10.20683/jniph.73.4_283

Many of the social issues that are currently being addressed require human involvement, making them difficult to solve with just a solution. For example, even if there is technology to monitor health status and a method to improve it, the actual subject may not take actions to monitor their own status, and even if they do monitor it, they may not take measures to improve it. Even in environmental issues and everyday life (urban development) issues that need to be solved to realize a sustainable society, no matter how advanced developments in recycling technologies and mechanisms become, it is important for residents to continue collecting waste in their daily lives in order to utilize them. In urban development efforts, the efforts of governments and companies alone are not enough, and it is essential for residents living in the area to be actively and continuously involved. In other words, people's proactive change of behavior, or "behavioral change," is extremely important for problem solving. For these reasons, in recent years, behavioral economics methods and approaches called "nudges," along with behavioral observation and marketing methods to understand consumers, have attracted attention as strategies and methods for promoting behavior change. For this reason, big data analysis and utilization of consumer purchasing behavior history data and internet browsing behavior history data are expected. While these have a long history in their own fields, with the recent spread of digital technology, the recording and accumulation of data, and the utilization of accumulated data through data science and artificial intelligence technology have become commonplace, and they can now be regarded as a common framework. In this paper, we introduce a probabilistic approach, Bayesian network, and probabilistic latent semantic analysis that handles "behavior change" within a framework in which human behavior is recorded and accumulated as data and which utilizes that data, as well as examples of applications of these methods to behavior change support technologies.