From Newton's laws to quantum mechanics, what are the key factors driving the
evolution of science? Science of Science (SciSci) is an emerging research field that
bears the mission to answer this question with scholarly data -- the digital trace of
scientific endeavor and influxes. Recent years have witnessed many exciting
advances, which have been documented in papers, patents, research grants and
awards, paper citations, scientist collaboration and mobility. These data are
becoming increasingly accessible in digital platforms, lending unprecedented
opportunities to establish mathematical models to quantify our understanding of
scientific evolution. Yet, scholarly big data has imposed new challenges for
traditional SciSci research, in the sense that they are:
1) heterogeneous and multi-modal (e.g., texts, figures, tables, equations),
2) with complex topological structures (e.g., scholarly knowledge networks),
3) continuously generated at a rapid pace.
Leveraging scholarly big data to answer key SciSci questions will require
methodologies beyond traditional statistical modeling.
Data Mining (DM) and artificial intelligence (AI), which have shown great potential
to uncover novel knowledge from such massive data, are thus expected to take a
crucial role at the frontier of SciSci research. Here we propose the AI4SciSci
workshop, focused on empirical findings, methodological papers, theoretical
underpinnings, and conceptual insights related to DM and AI in the broad research
field of SciSci. By connecting the research communities of DM, AI, and science of
science, this interdisciplinary event will not only introduce new data-centric tools
to answer questions in the SciSci research, but also inspire the development of
policy-relevant prediction tasks to push the frontiers of data mining research.
