Non-epistemic factors in epidemiological models. The case of mortality data

In this paper we unpack some of the roles that non-epistemic factors might have in the making of epidemiological models of COVID-19, first by a zoom-in on the definition of death due to COVID-19, then with a bird’s-eye view on different parameters that can be incorporated into predictive epidemiological models. Specifically, the definition of a death due to COVID-19 rests on a choice, the choice of positively evaluating the goal of fostering public health through disease prevention and treatment, over the goal of preserving the epistemic and biomedical soundness of a causal inference about the underlying cause of death. In its turn, this definition figures as a component of death-related parameters, such as infection fatality ratio and case fatality ratio. Death-related parameters, however, collectively represent only one option, as other measures may optionally be considered when modeling, such as potential years of life lost. Choosing between different parameters in epidemiological modeling does not depend on facts only, but also on goals and value assumptions, that is on non-epistemic factors. Arguing that non-epistemic factors play a key role not only in determining what counts as a death due to COVID-19 but also in realizing the epidemiological models of the current pandemic, we do not mean to claim that such mortality data and epidemiological models are flawed, unreliable, useless or non-purely scientific. Philosophers of science have recently provided convincing accounts of the role of values in science. Rather, we want to highlight that the interplay of facts and values must be made explicit to the general public and the two components must be carefully separated and evaluated independently. Whereas facts can only be acknowledged, values and goal-related choices may and should be rationally discussed on practical and ethical grounds.