Many decisions in modern societies have a very complex scientific basis. Clinicians have to choose between different drugs for treating a patient. Central bankers have to forecast the evolution of financial
markets, and to control the amount of money that circulates in a society. Physicists have to evaluate the impact of continued CO2 emissions for life on the planet. All these decisions are based on the
forecasts of scientific models, and sometimes, their predictions reach a great degree of exactness (e.g., in identifying high-risk hospital patients and allocating resources efficiently).
However, why should we rely on such models when they are highly idealized and contain assumptions that are far from the truth? What is it that makes decisions based on them reliable and trustworthy? How
do we factor in their intrinsic uncertainty? In short, how does science based on uncertain models contribute to good decisions?
Our project investigates the interface between modeling and decision-making. We develop an understanding of how scientific models function, how they advance our knowledge despite their intrinsic
uncertainty, and how they are interpreted in a decision context. More specifically, we focus on the following three questions, which correspond to our main targets:
1. How can highly idealized and intrinsically uncertain scientific models be successful in prediction?
2. Why can we trust and accept scientific models in spite of their intrinsic uncertainty and how should we factor in this uncertainty?
3. How should we synthesize actuarial, model-based judgment with human expertise in making practical decisions?
In answering these three questions, our projects integrates foundational philosophical analysis (e.g., rational criteria for theory acceptance), formal and conceptual analysis, and case studies about
construction and use of models in a number of relevant scientific disciplines like financial economics and evidence-based medicine.
The outcomes of our project explain the epistemic value of uncertain scientific models, and how they guide rational decisions. This is of utmost relevance in an age of science skepticism, where the authority
of scientists (and the model-based predictions they make) is often challenged by claims that models are intrinsically uncertain and hence the policies adopted on their basis are not trustworthy (e.g., global
warming, quantitative easing, and vaccination policies). The international reputation of our research team, its experience in interdisciplinary projects, and the collaborations with mathematicians, economists,
and medical scientists within the affiliated institutions, guarantee that our research objectives can be met and will substantially advance the state of the art.
The overall project is divided into three subprojects, each of which employs two postdoctoral researchers ("assegnisti di ricerca") and is coordinated by the leader of a local research unit. The subprojects
correspond to three different stages on the path from models to decisions: (1) the construction of idealized models and the evaluation of their intrinsic uncertainty and (mis)match with reality; (2) the
acceptance of a particular scientific model on a certain evidential basis; (3) the use of models in decision-making, including the cognitive pitfalls that arise when models represent uncertainty in probabilistic
terms, and the problem of synthesizing model-based judgment and human expertise. The project PI (J. Sprenger/UniTO) is in charge of coordinating the project, keeping the parts coherent and synthesizing
the results in a final book project.
In Subproject 1 (leader: G. Valente/PoliMi), we investigate how models are constructed in statistical physics and related disciplines with a strong decision component, such as climate science and
econophysics. These models typically exhibit a high degree of idealization. We identify where exactly they mismatch reality, and what kind of predictive ambitions t