MaLGa Colloquia - Cancer Screening: A Parable of Prediction
09 May 2023, 16:00 — Room 322, UniGe DIBRIS, Via Dodecaneso 35 You can attend remotely by joining the Teams call at the link below. Meeting ID: 337 626 222 094 Passcode: sTotE8
![IMG-5471 - [Fronte Naso Capelli]](/media/images/IMG-5471_jvgfr7_333YgLe.width-800.jpg)
Speaker:
Benjamin Recht — University of California, Berkeley
Benjamin Recht — University of California, Berkeley
Abstract:
The central promise of machine learning in medicine is improving predictions in healthcare settings. But this begs the question of the importance of prediction in medicine. This talk will examine the value of prediction in healthcare through the history of cancer screening. I will describe how researchers persuaded themselves that early detection would cure cancer, how the evidence generated for this belief was flawed, and how the early detection paradigm remains controversial 50 years after the first randomized screening trial. I will highlight lessons for machine learning regarding the quality of labels in medical detection problems, the dynamic distribution shifts arising in healthcare populations, and the broader role of predictions guiding medical interventions.
The central promise of machine learning in medicine is improving predictions in healthcare settings. But this begs the question of the importance of prediction in medicine. This talk will examine the value of prediction in healthcare through the history of cancer screening. I will describe how researchers persuaded themselves that early detection would cure cancer, how the evidence generated for this belief was flawed, and how the early detection paradigm remains controversial 50 years after the first randomized screening trial. I will highlight lessons for machine learning regarding the quality of labels in medical detection problems, the dynamic distribution shifts arising in healthcare populations, and the broader role of predictions guiding medical interventions.
Bio:
Benjamin Recht is a Professor in the Department of Electrical Engineering and Computer Sciences at the University of California, Berkeley. His research has focused on the application of mathematical optimization and statistics to problems in data analysis and machine learning. He is currently studying histories, methods, and theories of scientific validity and experimental design.
Benjamin Recht is a Professor in the Department of Electrical Engineering and Computer Sciences at the University of California, Berkeley. His research has focused on the application of mathematical optimization and statistics to problems in data analysis and machine learning. He is currently studying histories, methods, and theories of scientific validity and experimental design.
Links: