Innovation and Order in Citation Networks
Slides for talk given on Wednesday 12th July 2023 at NetSci 2023 in Vienna
Innovation and Order in Citation Networks
PRESENTER: Tim Evans
Based on work by Henry C W Price, Martin Ho, Eoin O'Sullivan and Tim Evans.
ABSTRACT. We study citation data on eight vaccines, including four COVID-19 vaccines, approved between 2013 and 2022 and based on one of four different vaccine platforms. Our data is obtained from four types of document: drug approvals, clinical trials, patents and journal articles. Our data comes from ClinicalTrials.gov, Lens.org and Dimensions.ai. Each type of document is a node in a distinct layer of the network. Starting from the approval document issued by the US Federal Drug and Food Administration (FDA) for one of our eight chosen vaccines, we follow the bibliographical references back for a number of steps to produce citation networks of between 12 and 113 thousand nodes and an average degree of around 14.5.
The sense of order encoded in a DAG allows us to assign a unique height h(v) and depth d(v) to every node v. We define the criticality c(v) of a node to be c(v)=H-h(v)-d(v) where H is the largest height in the network. Any node lying on the longest path in the DAG will have zero criticality, and nodes that lie on paths that are slightly shorter than the longest path will have small criticality values. Our conjecture is that nodes with low criticality are the most important documents for the innovation process. In geometric terms, our inspiration comes from the fact where a DAG is embedded in a Minkowski space-time, the longest path in the DAG is the closest path to the space-time geodesic, the path of least resistance, least action.
Our method gives us a path of events that narrates innovation bottlenecks. We quantify the position and proximity of documents to these innovation paths to identify key innovation events. We also have information on funders. We show that when it comes to vaccine innovation, diffusion-oriented entities are preoccupied with basic research; biopharmaceuticals tend to participate in applied research and development activities; while challenge-led entities tend to sit in the middle.
