[quercusa]

There's something quite like this in US law. If you are citing a case in a legal brief, you need to make sure it hasn't been overturned along the way, so you 'Shepardize' it. From Wikipedia [0]:

Shepard's Citations is a citator used in United States legal research that provides a list of all the authorities citing a particular case, statute, or other legal authority. The verb Shepardizing refers to the process of consulting Shepard's to see if a case has been overturned, reaffirmed, questioned, or cited by later cases.

[0]: https://en.wikipedia.org/wiki/Shepard%27s_Citations

[jmnicholson]

We're effectively bringing shepardizing to science at scite (scite.ai). Here's a piece I published on the topic recently: https://www.sciencedirect.com/science/article/pii/S259023852...

[timlod]

Hi! I checked out scite and really like the product- however, I can't seem to find relevant papers in my field (ML&AI), ie., I searched for the batchnorm paper, which there has been a lot of 'reevaluation' of- but it's not available. I suppose at this point you don't have access to (some of) the big AI journals/conferences. Is this something on the horizon?

[em-bee]

this sounds like a good approach. right now every citation is an endoresement, and the choice is to support a publication or to ignore it. there should be a way to make a negative citation, that allows you to declare that the referenced publication is contradicting. as a result there would be two citation counts for each publication. like upvotes and downvotes. i believe contradicting citations do happen now, but the fact that they are contradicting can't be seen from the citation count.

[d110af5ccf]

> right now every citation is an endoresement

This is a valid point, but it's important to note that it only applies when using aggregate citation counts as a rough proxy for importance within a field. That's useful for prioritizing what to read but not for assessing the validity of any given result (of which there are often quite a few, at least in the life sciences).

Within a given paper, it's not at all uncommon for the authors to explicitly call out some detail from another study as being incorrect in their view. That doesn't mean that they necessarily agree or disagree with the rest of the cited work though.

What I'm getting at here is that negative citations would likely be far too coarse to be useful in practice. It's relatively rare that a paper outright disagrees with an entire work.