Quantification of Scientific Discovery Process Implies Better Science


The ever-increasing amount of published science poses a challenge in interpretation and validation of these publications and the formation of scientific facts. Despite the apparent lack of alignment between published claims and established facts, accounting for network structure enables predictive models that can assess the validity of published claims and render scientific facts, using the example of gene-gene interaction data. Our simulations, based on pre-trained models, imply that the overall knowledge of facts can be improved by shifting the attention of the scientific community or modifying the funding policies. In conclusion, we review alternative approaches to optimization of the scientific discovery process and discuss alternative domains of application of our methodology.

Sep 9, 2022
Seagate AI/ML working group (remote)