Causal models, creativity, and diversity

Abstract

Causality applies everywhere, and it is hard even to imagine a world in which it does not. Yet, one must acknowledge that life also is creative and diverse. Un-der these circumstances, the question emerges whether causal models can ex-plain life's creativity and diversity. Some life scientists say yes, yet many hu-manities scholars cast doubts or have posited that they have reached the end of theory. Here, I build on common empirical observations as well as long-accumulated modeling experience, and I review and further develop a unified framework for causal modeling that applies to all sciences including physics, biology, the sciences of the city, and the humanities.

Neural Networks through the Lens of Evolutionary Dynamics

“…The two equations that this article has recovered from NNs have a history spanning centuries. As part of this history, they have found application in a truly breathtaking array of sciences (Baciu 2023).…”

“…Building on this body of knowledge as well as additional experiments in the humanities, I have argued that variation-selection processes can be used to describe creativity in any kind of physical, biological, social, or cultural system (Baciu 2023). A 2023-article gives an overview of creative transformations that can be described with this math.…”

“…Evidently, frequency dependent selection works to describe processes in an immensely broad array of applications. In the 2023-article just mentioned, I have made the case that frequency dependent selection can describe diversification and interplay in any kind of physical, biological, social, or cultural system (Baciu 2023). The rationale and empirical support are found in the article.…”

“…From the mathematical description of RNNs, I recover the quasispecies equation, which is a general equation of evolution, used to simulate variation-selection processes. This equation has been shown to describe evolution and creativity across a broad range of disciplines (Domingo & Schuster, 2016, Singh et al, 2023, Baciu 2023. Furthermore, from Self-attention layers, I recover the replicator equation, also known as generalized Lotka-Volterra equation, which is the general equation of evolutionary game theory (Hofbauer & Sigmund, 1998).…”

“…The replicator equation, specifically builds on more than a century of science. It has been used to describe processes of play and diversification across a broad range of disciplines (Lotka 1910, Ross 1911, Baciu 2023. Thus, the article recovers the equation of variationselection processes from RNNs and the equation of frequency dependent selection from Self-attention layers.…”