Getting Abstract Mathematical Models in Touch with Nature

Abstract

Argument

This article focuses on the scientific practice of a specific class of models in neuroscience and biology that approach biological systems as computational or information processing systems. This specific approach to biological systems has a long tradition that started with the information discourse in the 1940s. Borrowing concepts, methods, and techniques from cybernetics, information theory, and physics, these models are situated at the interface of different scientific disciplines. This article examines in detail the Hopfield model, a model of the specific brain function of auto-associative memory that is situated at the interface of neuroscience, theoretical physics, and engineering. By drawing analogies to a model of disordered magnetic systems in physics, John Hopfield constructed a model that was able to mimic auto-associative memory. The model became an active field of research in theoretical physics and in engineering but not in neuroscience. According to neuroscientists, in the process of construction the model had lost touch with the biological system and the function that it was supposed to model. How did this happen? As will be shown, the process of constructing the model was guided by conceptualizing the biological organism as a computational, information-processing device. This construction process gained its own momentum that involved an interrelated development of a computational concept and a redescription of the biological system or phenomenon for bringing the computational concept to work. We will see how physicists and engineers tried to get models in touch with biological systems by constructing “a synthetic model.”