2023, Biostatistics, Epidemiology and Public Health – Journal Club – “Estimating categorical counterfactuals via deep twin networks”

You are invited to participate at the Journal Club organized by the Unit of Biostatistics, Epidemiology and Public Health on the theme: “Estimating categorical counterfactuals via deep twin networks” from 1:30 p.m until 2:30 p.m.

The seminar will be presented by Dr. Andrea Pedot – University of Padova, and it will be moderated by Prof. Annibale Biggeri – Full Professor in Medical Statistics – University of Padova.

To register for the seminar, just click on the following link, register through the button “register” and then you will get the link to connect: https://events.teams.microsoft.com/event/5d887203-aa02-4fa4-9c14-62f0c7589890@bd5051ae-6f73-49ef-8068-9c057c23955a 

Abstract: “If my credit score had been better, would I have been approved for this loan?”, “What is the effect of the diabetes type on the risk of stroke?”. Scientists and the public alike routinely ask causal questions like these. Counterfactual inference is a powerful tool, capable of solving challenging problems in high-profile sectors. To perform counterfactual inference, one requires knowledge of the underlying causal mechanisms. However, causal mechanisms cannot be uniquely determined from observations and interventions alone. This raises the question of how to choose the causal mechanisms so that resulting counterfactual inference is trustworthy in a given domain. To learn such causal mechanisms, and perform counterfactual inference with them, Balke and Pearl introduced deep twin networks in 1994. These are deep neural networks that, when trained, are capable of twin network counterfactual inference—an alternative to the abduction, action, & prediction method of counterfactual inference”.