Mapping interdependencies among the UN's Sustainable Development Goals via semantic overlap in stated targets and current progress
Thomas Bryan Smith, Raffaele Vacca, Ilaria CapuaAt the 2015 United Nations (UN) General Assembly the 193 member states of the UN signed off on the 2030 Agenda for Sustainable Development. This agenda included 17 Sustainable Development Goals (SDGs), developed at the Rio+20 conference in collaboration with multiple stakeholders, and intended to spur on improvements in economic growth, environmental sustainability, global health, and peace. The 17 SDGs are both heterogeneous and partially overlapping, each of them comprising different sub-targets (169 individual targets overall). This internal heterogeneity produces a vast and complex set of interdependencies among the SDGs and the research and policies that may contribute to them. Some SDG clusters may be better addressed simultaneously, while others may be marred by necessary trade-offs. We use topic models, including both Latent Semantic Indexing and Latent Dirichlet Allocation, to analyze a large body of text describing SDG ‘targets & indicators’ and annual ‘progress & info’, which was web-scraped from the official UN’s Sustainable Development Goals Knowledge Platform. Results from these topic models are then used to generate networks mapping semantic links and overlaps among the SDGs, including edge overlap scores and edge and node keywords. We apply different techniques of network and cluster analysis to identify groups of SDGs which, based upon their semantic overlap, could be addressed simultaneously, with edge keywords providing qualitative insights. Results confirm patent overlaps (e.g. Economic Growth and Consumption/Production; Climate Action and Live Below Water), and unveil subtle overlaps (e.g. Peace/Justice, Poverty). In a second part of the study, we incorporate data on titles and abstracts of publications at a large research university to evaluate the extent to which collaborative networks of scientists, and the research they produce, align with or address the SDGs and their sub-targets. This project illustrates the advantages and limitations of topic modeling and network analysis methods as applied to the study of scientific production and collaboration in a specific interdisciplinary research field and policy area – sustainable development in our case. Implications, limitations, and directions for future research are discussed.