Periodic Reporting for period 2 - PREDICT (The Future of Prediction: The Social Consequences of Algorithmic Forecast in Insurance, Medicine and Policing)
Période du rapport: 2021-08-01 au 2023-01-31
The algorithmic turn of prediction, connected with big data and machine learning, presents an exciting and urgent challenge for the social sciences. We are facing a radical technical-computational innovation, whose effects extend throughout society, promising innovative solutions and yet producing disruptive problems.
To address these challenges, the project carries out a set of theory-driven empirical studies of the transition from probabilistic forms of uncertainty management to the new algorithmic forms of prediction in three important social areas – insurance, medicine and policing.
In the field of insurance, Insurtech promises precise and individualized risk forecasting, that should allow companies to know the specific needs and characteristics of their customers and to propose personalized offers, messages, pricing and recommendations. The problem, however, is that individualization could undermine the mutualization principle on which insurance is based, with the risk of inefficiency and discrimination, and even of emptying its business model.
In the medical field, Precision Medicine promises to turn healthcare from a reactive to a proactive field by enabling the prediction of future needs of patients and the customization of medical interventions according to such highly specific predictions. This should allow to overcome the wastes and inefficiencies inherent in the current "one-size-fits-all" approach. The problem is that personalization must be combined with established approaches and methods in medical practice (both in the clinic and in research), that address large populations - with inevitable consequences for the experience of patients and staff, as well as for scientific activity and institutions.
In the field of crime prevention, algorithmic prediction has led to the development of Predictive Policing, applying digital techniques that claim to identify potential criminal activities and to intervene before they are accomplished. Predictive methods are expected to allow police to move from a reactive to a proactive approach. But this attitude does not take into account a fundamental distinction that our research is addressing: that prediction by itself is not prevention. Our analysis is focuses on the relationship between prediction and prevention and on its consequences for policing activity.
To address these challenges, the project carries out a set of theory-driven empirical studies of the transition from probabilistic forms of uncertainty management to the new algorithmic forms of prediction in three important social areas – insurance, medicine and policing.
In the field of insurance, Insurtech promises precise and individualized risk forecasting, that should allow companies to know the specific needs and characteristics of their customers and to propose personalized offers, messages, pricing and recommendations. The problem, however, is that individualization could undermine the mutualization principle on which insurance is based, with the risk of inefficiency and discrimination, and even of emptying its business model.
In the medical field, Precision Medicine promises to turn healthcare from a reactive to a proactive field by enabling the prediction of future needs of patients and the customization of medical interventions according to such highly specific predictions. This should allow to overcome the wastes and inefficiencies inherent in the current "one-size-fits-all" approach. The problem is that personalization must be combined with established approaches and methods in medical practice (both in the clinic and in research), that address large populations - with inevitable consequences for the experience of patients and staff, as well as for scientific activity and institutions.
In the field of crime prevention, algorithmic prediction has led to the development of Predictive Policing, applying digital techniques that claim to identify potential criminal activities and to intervene before they are accomplished. Predictive methods are expected to allow police to move from a reactive to a proactive approach. But this attitude does not take into account a fundamental distinction that our research is addressing: that prediction by itself is not prevention. Our analysis is focuses on the relationship between prediction and prevention and on its consequences for policing activity.
The PI has developed her background research on the theoretical prerequisites of the social use of algorithms, which has produced a volume published in English at MIT (being translated into Italian by Bocconi University Press), 5 articles in international peer reviewed journals and 3 book chapters.
In the field of insurance, we established a network of contacts with the main car insurance companies and international software providers, an ongoing collaboration with Swiss Re and the "Programme de recherché sur l’Apprehension des Risques et des Incertitudes" (PARI) in Paris, carried out a series of meeting and interviews, participated in 7 international conferences and meetings, organized a workshop in Bologna in November 2021. We produced 2 articles in international peer reviewed journals and 4 articles in relevant insurance journals.
In the field of medicine, we carried out a quantitative analysis of a dataset of articles in the years 2000 to 2019 and analyzed them with structural topic modeling (STM). This analysis led us to suppose that a new separation between clinical epidemiology and population epidemiology is emerging after the approximation and methodological uniformity in the last decades of 1900. The article presenting our research was accepted for an oral presentation at the XLVI Convegno Associazione Italiana di Epidemiologia, and will be now submitted to a sociological journal (extended version) and to an epidemiological one (shorter version). We also conducted an in-depth preliminary study and ethnographic observation of Molecular Tumor Boards, which are the meeting place of traditional to digital approaches in the medical field, and established a series of contacts to carry out ethnographic research in the coming months.
In the field of policing, we intensified our collaboration with the German Police University in Muenster with the partial affiliation of one of our researchers, we applied for and obtained permission from the Ministry of the Interior of the Land Nordrhein-Westfalen to conduct our research, and began a collaboration with the team of the Project SKALA that produces the software. We participated in 14 international conferences and meetings, organized an international workshop and co-organized 2 additional ones. We produced 2 articles in international peer reviewed journals, 4 book chapters, co-edited a collective book and co-authored another one.
In the field of insurance, we established a network of contacts with the main car insurance companies and international software providers, an ongoing collaboration with Swiss Re and the "Programme de recherché sur l’Apprehension des Risques et des Incertitudes" (PARI) in Paris, carried out a series of meeting and interviews, participated in 7 international conferences and meetings, organized a workshop in Bologna in November 2021. We produced 2 articles in international peer reviewed journals and 4 articles in relevant insurance journals.
In the field of medicine, we carried out a quantitative analysis of a dataset of articles in the years 2000 to 2019 and analyzed them with structural topic modeling (STM). This analysis led us to suppose that a new separation between clinical epidemiology and population epidemiology is emerging after the approximation and methodological uniformity in the last decades of 1900. The article presenting our research was accepted for an oral presentation at the XLVI Convegno Associazione Italiana di Epidemiologia, and will be now submitted to a sociological journal (extended version) and to an epidemiological one (shorter version). We also conducted an in-depth preliminary study and ethnographic observation of Molecular Tumor Boards, which are the meeting place of traditional to digital approaches in the medical field, and established a series of contacts to carry out ethnographic research in the coming months.
In the field of policing, we intensified our collaboration with the German Police University in Muenster with the partial affiliation of one of our researchers, we applied for and obtained permission from the Ministry of the Interior of the Land Nordrhein-Westfalen to conduct our research, and began a collaboration with the team of the Project SKALA that produces the software. We participated in 14 international conferences and meetings, organized an international workshop and co-organized 2 additional ones. We produced 2 articles in international peer reviewed journals, 4 book chapters, co-edited a collective book and co-authored another one.
In all areas of the project, our research has shown the emergence of novel ways of combining probabilistic and algorithmic forms of prediction, which take on complex and diverse configurations in different areas. This finding goes far beyond the standard juxtaposition between generalized forecasting and innovative forms of personalized forecasting, and promises to be extremely useful in concretely analyzing ongoing developments.