Epistemic Rights and Responsibilities of Digital Simulacra for Biomedicine

Enabling the Nonhypothesis-Driven Approach: On Data Minimalization, Bias, and the Integration of Data Science in Medical Research and Practice

“…A typical example concerns (the effects of) the decision to use dimension-reduction techniques (Chung et al 2021). Note, however, that such techniques do not necessitate the type of simplification that Cho and Martinez-Martin describe because "key feature" selection is not hypothesis-driven (Cho and Martinez-Martin 2023). Subsequently, contrary to their claims, such models retain the potential to detect unexpected patterns in data or generate surprising results (Chung et al 2021;Stingone et al 2021).…”

“…While Cho and Martinez-Martin recognize ethical benefits of "digital simulacra," such as speeding innovation, lowering the costs of innovation, and minimizing risks to human beings via in silico drug testing, they claim that data minimalization is a looming barrier (Cho and Martinez-Martin 2023). They argue that "regulations that oblige researchers to collect on the minimum necessary protected health information conflict with the analytic needs of digital simulacra developers" because these developers "attempt to collect as much data as possible" (Cho and Martinez-Martin 2023, 49).…”

“…In effect, Cho and Martinez-Martin deny that the nonhypothesis-driven approach has the potential to reduce or eliminate biases that originate from researchers (Cho and Martinez-Martin 2023). They write that this approach is "depicted" as being able to do so, but argue that "In practice, however, features of digital simulacra have the potential to increase bias, obscuring values and inequities that are embedded in the decisions made throughout the design process" (Cho and Martinez-Martin 2023, 47).…”

“…Cho and Martinez-Martin provide a wide-ranging analysis of what they label "digital simulacra"-which are in essence data-driven AI-based simulation models such as digital twins or models used for in silico trials-that explores many ways in which "digital simulacra" could affect certain (perceived) ethical and epistemic values (Cho and Martinez-Martin 2023).…”

“…Their analysis outlines challenges and limitations of incorporating "digital simulacra" in healthcare, such that potential harms can be mitigated (Cho and Martinez-Martin 2023). While their analysis provides a starting point for understanding the ethical implications of such models, it is our central contention that their analysis misses an identification of the way in which data are selected through the "data-first" or nonhypothesis-driven approach.…”