Models are indispensable for public policy making. When faced with complex questions, policy makers often turn to scientists to synthesize information, conduct analyses, and provide predictions that they can use as a basis for making difficult decisions. What happens when these two worlds meet, however, can be anything but straightforward. Differences in expectations, motivations, values, and scientific literacy between scientists and decision makers can lead to misunderstanding, while social, cultural, and political considerations can influence how even the best science-based models are interpreted and used.
Dec 19, 2024
Events
HLRS Projects
Simulation
AI & Data Analytics
Philosophy & Ethics
See all news
Ideally, models would lead to clear, easily executed recommendations. However, developing models of complex problems is often not a simple process. Factors including the ethical values that motivate a modeling approach, practical considerations such as what datasets are available, and the selection of which parameters to prioritize within a complex system can greatly determine a model’s results. On the policy side, considerations such as sociopolitical goals, the ethics of using models, and the feasibility of implementing model-based recommendations can also determine how they are used in practice. Such matters have become more relevant with the advent of more powerful “black box” applications of computer simulation and artificial intelligence, while questions of how to manage increased skepticism of modelling approaches outside the scientific community have become more urgent since the COVID-19 pandemic.
On November 25-27, 2024 the Department of Philosophy of Computational Sciences at the High-Performance Computing Center Stuttgart (HLRS) organized and hosted an international conference titled Modeling for Policy. Bringing together a multidisciplinary group that included philosophers, social scientists, historians of science, and users of computing technologies for modeling, the event offered the opportunity to reflect on the practice of modeling and the challenges it can face in the context of public policy making. This latest event in the conference series Science and Art of Simulation covered a wide range of domains in which models are used, including public health, energy transformation, climate modeling, employment programs, and value chain analysis, among others. Within these contexts the event sought to define the capabilities and limits of simulation more clearly, explore how trust in models could be improved among policy makers and the general public, and make recommendations that could lead to better interactions between simulation scientists and society at large.
In a keynote address, for example, public health researcher Stephanie Harvard of the University of British Columbia explored how values inform the development of computational models. Considering a case study concerning the potential effectiveness of HEPA filters to counteract air pollution caused by forest fires in Canada, she advocated for a more self-reflective approach on the part of modelers. While showing that no strategy for managing values is perfect when considered from a philosophical perspective, she suggested that modelers have an ethical responsibility to identify and articulate the values and judgments that inform their work. Such values management should include being clear about what outcomes are considered desirable or undesirable in a policy context, what pivotal decisions were made in the development of the model (such as what data to include or exclude), what applications of the model are appropriate, and who is accountable for its results. Transparency about such issues is important particularly in a public policy context, where nonexperts must often make decisions based on specialized research they might not fully understand.
The capabilities of simulation have progressed as computing power has grown, and scientists now have the ability to generate increasingly large and detailed models. As biostatistician and health policy specialist Alyssa Bilinski of Brown University argued in a second keynote address, however, simply building more complex models is not always the most effective way for scientists to advise public policy makers. Instead, she argued for the utility of what she calls “napkin math,” an approach to analytical thinking that starts with the simplest possible model. In some cases, she suggested, even a basic-level analysis is sufficient to provide practical answers to complex questions, and can be produced faster and in a way that is easier to explain than large-scale simulations. In addition, modelers can use napkin models to check the coherence and feasibility of more complex models, or to interpret and compare models. Bilinski also pointed out that an awareness of what models can and cannot do in the context of policy making is important, remarking, “Good policy modeling starts a conversation, rather than ending it.”
This year’s Science and Art of Simulation conference also benefited from a partnership with the Gesellschaft für Wissenschaftsforschung, a Berlin-based international membership association that holds meetings and publishes scientific papers focused on the study of science and its methods. In two sessions organized by the society, including a keynote lecture by University of Tübingen philosopher and historian of science Reinhard Kahle, speakers focused on the opportunities and risks of the increasing use of artificial intelligence and large language models in scientific research and in education.
Throughout the three-day conference, speakers expressed confidence that scientific modeling offers valuable tools for public policy makers. At the same time, however, recent events show why scientists can’t simply assume that scientific “truth” will alone automatically determine policy making. Modeling for public policy is itself embedded in a complex space that is influenced by constellations of social, political, economic, and cultural factors. As the conference showed, a more complete understanding of the relationships among these domains will be needed to maintain public support of good science and to ensure that it continues to drive public policy. Interdisciplinary research like that facilitated by HLRS’s Department of Philosophy of Computational Sciences and its SAS conference series can help to understand the broader context for science, and to identify strategies for improving the integration of scientific thinking and modelling into public decision making.
— Christopher Williams
SAS24 Conference program and abstracts
HLRS Department of Philosophy of Computational Sciences
