Policy Twin: Toward evidence-based policy-making — Maximizing policies’ impact through digital space verification

To help address complex social issues such as climate change, economic disparities, and a declining and aging population, Fujitsu has been advancing research and development in converging technologies that combine humanities and social sciences with computer science. As part of these activities, Fujitsu is developing Social Digital Twin (*1), which recreates real-world human behavior and social dynamics in a digital space, enabling the effectiveness and impact of policies to be tested and verified in advance of implementation.

Building on this foundation, Fujitsu has advanced this concept further, with the introduction of our new technology: Policy Twin. This is a technology that goes beyond simply representing the state of people and society, and actually expresses policies themselves (*2) —implemented across various domains—in machine-readable flow formats that can be handled within a digital space. This makes it possible to formalize policies as explicit knowledge and apply a wide range of mathematical and analytical processes. One of the key features of Policy Twin is its ability to derive diverse policy options automatically while taking social acceptability into account.

In this article, we present a case study that demonstrates the value that Policy Twin can deliver, in this instance by applying it to municipal initiatives for preventing the progression of diabetic nephropathy. As well as setting out the value it delivers to users, the case study also features behind-the-scenes stories about its development.

── Tell us some more about the technical validation in a diabetes nephropathy progression prevention program.

Sayuri: Diabetic nephropathy is a disease that may progress to the point where dialysis becomes necessary, significantly impairing patients’ quality of life. It also contributes to rising medical costs at the national level. Against this backdrop, municipalities across Japan have been promoting initiatives aimed at preventing the progression of diabetic nephropathy by identifying high-risk individuals at an early stage and providing appropriate health guidance. In this project, we collaborated with a municipality to apply our technology to initiatives focused on preventing the progression of diabetic nephropathy.

── What do you need to take into consideration when identifying high-risk individuals?

Sayuri: When identifying high-risk individuals, it is essential to base the approach on national guidelines, while also taking into account the balance between a municipality’s operational capacity and costs. The number of people identified can vary significantly depending on the selection criteria, which in turn affects the amount of manpower and budget required. For example, even something as simple as how high-risk individuals are contacted—whether by phone, mail, or other means—differs from one municipality to another. As a result, initiatives that have proven effective in other municipalities may not necessarily be feasible under the organizational structures or budget constraints of one’s own municipality. Therefore, policies must be designed with careful consideration of both effectiveness and practicality, taking into account the specific circumstances of each municipality.

── How did you utilize Policy Twin?

Sayuri: First, based on the municipality’s policy documents related to measures for preventing the progression of diabetic nephropathy, we created policy flows that represented those initiatives. At the same time, we also developed policy flows for similar initiatives being implemented in other municipalities. We then decomposed the large number of policy flows we had created into their constituent components and reassembled them to generate new policy candidates automatically. For each candidate, we conducted simulations covering factors such as medical cost reductions, health improvement effects, and the costs associated with implementing outreach notifications. As a result, we were able to confirm that it is possible to derive policy candidates that are expected to reduce medical expenses and improve health indicators through health guidance. This could be achieved while still satisfying constraint conditions such as the available human resources required to provide that guidance.

── What value does Policy Twin offer to its users?

Yuta: One of the major values of Policy Twin lies in digitizing policies in the form of flows, which makes it possible to view, in a mathematical and structural manner, content that previously could only be understood as individual documents. This enables users, for example, to make quantitative comparisons with similar or successful cases from other municipalities more easily, allowing policy discussions to move forward without excessive reliance on experience or established practices. In addition, another important value of Policy Twin is its ability to help users formulate new policy candidates in a way that fosters understanding and facilitates implementation. As policy candidates newly generated by Policy Twin are flexibly reconfigured based on existing policies, they incorporate new perspectives while remaining consistent with past initiatives and on-the-ground knowledge and experience. This makes them easier to accept in practical settings. This capability—deriving new policies while maintaining continuity with existing ones—is one of the key strengths of the Policy Twin algorithm. For more detailed information, please refer to our TechBlog (*3).

── Can this technology be applied to other fields?

Sayuri: Yes. Let me introduce some of the domains we have in mind and examples of potential applications.

— How was the Policy Twin development team formed?

Yuta: The Policy Twin research and development team is composed of members with diverse expertise, unified around the concept of converging technologies—integrating different scientific and technological disciplines to address social challenges. What makes the team distinctive is not only the fact that it brings together specialists from a wide range of technical fields such as speech processing, image processing, data mining, and robotics, but also that it includes researchers specializing in the humanities and social sciences, including economics.

I myself have been involved in research and development related to computing, but I was strongly drawn to this cross-disciplinary approach and joined the team. After becoming part of the team, I found that this interdisciplinary nature naturally manifests itself in our day-to-day research activities. For example, during our regular sessions where we introduce and discuss academic papers, even when we are addressing the same theme, the perspectives differ, and new papers and unique viewpoints are brought to the table every time. These diverse perspectives play a vital role in deepening discussions and refining our research themes.

— Could you tell us about your approach to conducting research?

Sayuri: We have advanced our research with a strong focus on listening to voices from the field. At the initial stage of the research, we visited people involved in policy-making and conducted interviews to understand on-the-ground challenges, based on which we developed hypotheses and technical concepts. In order to gain new perspectives, we also spoke with professionals involved in game design. The methodologies used to design virtual worlds share many commonalities with policy-making, which is essentially the design of the real world, and these insights proved very valuable. As the research progressed, we continuously refined the technology by incorporating feedback from those with practical, on-the-ground experience, gradually brushing up the technology step by step.

Yuta: We are engaged in cutting-edge research, and to accelerate our progress, we are also conducting joint research with top-tier universities in Japan and abroad (*4). In working toward solutions for a wide range of social issues, we closely collaborate with university researchers on topics such as the development of new mathematical modeling technologies and their real-world implementation. At times, we visit universities to give research presentations or take part in discussions, and when I returned to my alma mater after some time, I felt a sense of nostalgia. At the same time, it made me reflect on my current position—engaging with universities through research in a very different way from when I was a student.