Using data for risk management policy - introducing the data-oriented approach into policy-making in Japan on food safety, drug safety, earthquake disaster prevention, and climate change

Managing risks for the public is an important part of governments’ role. For a broad range of risks – chemicals in environment and food, nuclear accidents and natural disasters, epidemics, climate change, and so on – governments and international organizations perform risk management. Recently, the ways in which data are used for risk management in these policy fields have been changing rapidly.

This paper examines how the data-oriented approach has been introduced into the process of risk management policy-making in Japan. Here, the word “data-oriented approach” is counterposed to the “mechanism-oriented approach”. While these two should not be understood in terms of dichotomy, the former is inclined to value generation of useful knowledge through algorithmic manipulation while the latter tends to value understanding of phenomena themselves.

              Many policy fields have been rather tardy in incorporating the data-oriented approach into the process of risk management. Usable data are abundant, along with advanced analytical techniques, but they are often underutilized. For example, QSAR (Quantitative Structure-Activity Relationship) tools, which can algorithmically predict biological activities of chemicals for which animal experiment data are not available, have existed for decades but only recently began to be employed in the process of approving the use of food additives. Why is that? This paper seeks to identify multiple technical and non-technical factors which work for and against the adoption of the data-oriented approach in risk management, by comparing multiple fields: food safety, drug safety, earthquake, and climate change.

              Availability of high-quality data and reliability of analytical models are crucial conditions for the data-oriented approach to be of use in policy-making. These conditions differ from one policy field to another. Each field is constantly seeking solutions to such questions as: what is the best trade-off between the quality and quantity of data; how and to what extent data can be extrapolated; and how to reconcile or integrate different models. These are always tough questions, but the Japanese authorities are now tackling them in cooperation with scientists and the international community.

              Since the data-oriented approach is still of limited use in many policy fields, the role that the mechanism-oriented approach should play remains large. The main question of this paper is how the two approaches can collaborate best? Although each policy field differs in terms of data availability, model dependability, and other technical and non-technical factors, this paper suggests that each field should strategically overcome the challenges related to the collaboration of the two approaches.

Research methodology and data used:

The purpose of this research is to generate an overall picture of ongoing integration of the data-oriented approach into risk management policy-making. We have chosen four policy fields for comparison: food safety, drug safety, earthquake disaster prevention, and climate change. Such comparison is effective because it would reveal common challenges and potential solutions relative to the introduction of the data-oriented approach into risk management policy-making. With that in mind, we have conducted interviews with several researchers and policy-makers for each of the four fields. We have also constantly surveyed literature, which is evolving at a fast pace.

Key findings:

Factors driving or hindering incorporation of the data-oriented approach into risk management policy-making are diverse. For example, rising opposition to animal experiment has been a significant driving force for introducing algorithmic tools such as QSAR into food safety processes. On the other hand, in drug regulation for example, risk of litigation has been a major fetter. In earthquake and climate change, academic culture and tradition of these fields, valuing and rewarding understanding of seismological and meteorological mechanisms, seems to have mitigated zeal for the data-oriented approach.

              Overall, the data-oriented approach appears to have large potential to be positively pursued in fields where:

• predicting capability of the mechanism-oriented approach is limited (e,g, theoretically estimating biological activity of new chemicals);
• practical, clear benefits resulting from policy actions rather than rationales and processes of policy-making are valued (e.g. forecasting earthquakes accurately and optimizing countermeasures);
• people are not so stringent about the responsibility and liability for policy consequences (e.g. predicting earthquakes and taking precautionary actions).

              But even in fields where the data-oriented approach has large potential, data availability and model dependability are parameters that can work for as well as against it. For example, in food safety, one could increase the amount of data used by QSAR tools by including data from academic papers in clinical medicine. But the quality of academic papers is generally lower than that of toxicological data gained through animal experiments which are based on rigorous protocols. Thus, to what extent one should include such data becomes a matter of tradeoff. Meanwhile, in climate change, variability of models developed by researchers of many countries is quite large. Efforts have been made to integrate or reconcile those models effectively and legitimately, but it is still difficult to accurately assess and narrow down uncertainties of those models.

              While opportunities and challenges of the data-oriented approach differ from one policy field to another, one thing seems to be common: The collaboration between the data-oriented approach and the mechanism-oriented approach is essential in realizing robust risk management. The two approaches are complementary in all policy fields, although the division of roles and the modes of collaboration between them would vary. Although cultural differences that often exist between data-oriented experts and mechanism-oriented experts can be a challenge, both sides should strive for mutual understanding and productive collaboration.