Abstract: Investigating the mechanism of knowledge flow from science to technology helps understand how scientific progress drives technological innovation. Thus, this paper first constructed a "science-technology" knowledge transfer network composed of keyword citation. Then, using exponential random graph models, we integrated knowledge attributes with the knowledge transfer process in a modeling approach that simultaneously considered endogenous network structures. Finally, we conducted an empirical analysis based on scientific papers and patent data in the gene editing field from 1990 to 2018. We find that the high economic value of scientific and technological knowledge inhibits knowledge transfer, as rational actors tend to engage in exploitative innovation based on existing high-value knowledge. However, the convergence of economic value facilitates the transfer process by helping to reduce transfer barriers through moderate cognitive distance. The academic value of knowledge contributes to advancing the knowledge transfer process, but this effect is not statistically significant. Under the influence of homogeneity effects, knowledge novelty and geographic proximity have a positive impact on the formation of knowledge transfer relationships from science to technology. Meanwhile, comparison with random networks demonstrates that the citation behavior of technological knowledge toward scientific knowledge may not be influenced by semantic proximity or knowledge potential. These results demonstrate consistency across knowledge network simulation models in different time periods.

Key words: Knowledge transfer network, Exponential random graph model, Technology connection, Keyword citations, Endogenous network structure