Pierre Luc Germain

The evaluation of biomedical models in vitro and in vivo models of diseases raises several methodological issues, and has generally been hampered by a lack of attention to the precise functions played by biomedical models. For if biomedical models are ultimately expected to inform us about human pathologies, they seldom do so in isolation, and get there through a wide variety of ways. An epistemological understanding of this process is therefore a precondition for their evaluation, and this thesis is an attempt at building such an epistemology.
Several of the examples used come from cancer research, especially xenograft models and models used in the context of large-scale drug screenings. Another important set of examples come in vitro models, with a particular focus on disease modeling using induced pluripotent stem cells.
I argue that the notion of model, if conceived as to apply to biomedical models, conflates into that of experimental system. I therefore propose an account of biomedical models that does not presuppose a fundamental divide between modeling and experimentation. I show that biomedical models are not simply scaled-down versions of their target, but instead projections of their target in a different space of representation. I argue for an instrumental role of biomedical models, and use this role to explore the diversity of proximal functions fulfilled by biomedical models. I propose the notion of distributed modeling to draw attention to the relations between model systems, and illustrate this by analyzing the interplay between in vitro and in vivo models. Finally, I explore the implications of this account for the evaluation of biomedical models, and more broadly for the topic of scientific representation.