scite is a deep learning platform that allows anyone to learn how a scientific paper has been cited, not just how many times and, in particular, whether it has been supported or contradicted. To do so, we have extracted and classified 430M citation contexts from 13M articles and are processing more daily.

The novelty and the scale of this dataset allow us to finally begin evaluating the quality of citations journals receive, not just the quantity. As a first step, we visualized the incidence of supporting and contradicting citation statements across nineteen different journals.

We found that the journals do differ several-fold in the incidence of supporting and contradicting citations.

The question then is: why do they differ?

What do you think? What leads to different distributions of citation contexts at the journal level? Let us know on Twitter or by replying to this post!