Combinatorial novelty
- Thagard2010creative
- Sreenivasan2013
- Uzzi2013atypical
- Kaplan2014double edged
- Youn2015
- Kim2015technological
- Wang2016bias
- Solé2016singularities
- Shi2023surprising
Shi2023surprising proposes a Hypergraph generative model for the set of “tags” (e.g., MeSH terms) for each paper and then uses this model to evaluate the degree of surprise that each paper brings. If a paper combines a highly unlikely combination of tags, then the paper is deemed to be surprising. The paper then argues that the surprise is a useful predictor of outsized impact and often occurs as a result of field-crossing studies.