In my opinion a lot of comments here are too hung up on the fact that the researchers used data collected by Buzzfeed. I'm not a fan of clickbait either, but th

In my opinion a lot of comments here are too hung up on the fact that the researchers used data collected by Buzzfeed. I'm not a fan of clickbait either, but their corpus is publicly available here: https://docs.google.com/spreadsheets/d/1ysnzawW6pDGBEqbXqeYuzWa7Rx2mQUip6CXUUUk4jIk/edit#gid=1756764129 https://docs.google.com/spreadsheets/d/1ysnzawW6pDGBEqbXqeYu... So as far as I'm concerned, so long as the authors acknowledge their source (which they did) then there's nothing wrong here. Even FiveThirtyEight builds most of their models from datasets obtained from third parties. A lot of data science research is carried out using public datasets because collection is often an expensive or time-consuming endeavor, especially for social science research. The authors even say in their acknowledgement "We thank Craig Silverman, Lauren Strapagiel, Hamza Shaban, Ellie Hall, and Jeremy Singer-Vine from BuzzFeed for making their data available, enabling our research".