How do natural sciences learn from social sciences? Investigating the prevalence, origin, and location of references

This paper examines how social sciences knowledge affects natural sciences by empirically investigating the prevalence, origin, and location of academic references to social sciences from a set of natural sciences publications. Results show that social sciences knowledge accounts for 2.9% of the knowledge base for natural sciences and influences 23.1% of publications in 2018. Psychology, Economics & Business, and Media & Communications contribute the majority of knowledge. Most social sciences disciplines are primarily cited in the Introduction and Discussion section, except for Economics & Business, Other social sciences, and Sociology exhibiting a higher presence in Methodology. The discovered growing presence and various functions of social sciences knowledge in natural sciences deliver several political implications for education, research evaluation, and science policy. For instance, college educators should provide students from natural sciences with more necessary trainings in relevant social sciences disciplines to help them navigate the cognitive and practical discrepancies between the two, such as epistemology and methodology. Publishers from natural sciences should be prepared to expand the reviewer pool with social scientists when necessary to avoid defective quality control on the utilization of social sciences knowledge in submissions. Funding agencies need to better understand the role and significance of social sciences knowledge in the practice of natural sciences research and provide ample support to encourage its dissemination and application.

The study has several limitations. The dataset we employ mostly covers publications from life sciences, omitting some other natural sciences disciplines such as physics. Additionally, the context of learned social sciences knowledge is not thoroughly investigated. In the future study, we will improve this study by expanding datasets and conducting topic modelling or keyword extractions to deliver nuanced and contextual analyses.