Abstract:
I will present an analysis of evolution of technology based on the database of all issued US patents from 1978 - 2012. For this data, patents are a proxy for technological innovations, and citations are a proxy for genealogical connections. There are several differences between this version of technological evolution and biological evolution, e.g. (i) evolution of technology implicitly involves human intent and will, and (ii) the patent citations create a genealogical network that is highly multi-parental, and so gives rise to highly reticulate genealogical networks. Despite these differences from the biological context, I will argue that it makes sense to ask if technological change is in fact an evolutionary process. I will show how 'term frequency--inverse document frequency' (TFIDF) and word vector statistics may be used as a proxy for traits, and analyze a multi-parental version of the Price equation for evidence of evolution. We will see technology does seem to be an evolutionary process, and that particular technological traits display intermittent evolvability. Parallels and dissimilarities with biological and chemical evolution will be discussed, with a particular eye on the possibilities of data analysis for those systems.