7

Altmetric is a service that attempts to track the online impact of scholarly articles, which it does by keeping score of mentions on Wikipedia, Facebook, Twitter, blogs, and the like and then using questionable cutting-edge statistical methods for producing one shiny number for bean-counting administrators proud universities to list as part of their impact. On individual papers, it's the little multicolored rosette that looks like this:

Example of an Altmetric rosette

As part of their sources, Altmetric claims to include, among others, quote

– Q&A (stack overflow)

unquote. This is definitely the case: as an example, clicking on the rosette thingammy on this paper leads here, and rightfully shows a mention on this post on Physics Stack Exchange.

Questionable numbers aside, this service does have the potential, if it becomes reliable enough, for authors and readers to find people and places that have found some bit of research useful, and to explore those in more tailored and qualitative ways. As regards Stack Exchange, for example, the Physics site has 5k+ posts that reference either DOIs or the arXiv, and I'm sure that (i) many of those authors would be glad to hear of it, and more importantly (ii) the site would often benefit if they managed to find their way here and provided input or perspectives.

However, I'm sometimes confused about how the scraping works, and it often feels inconsistent to me. As an example, this Altmetric page turns up this answer but not this question, even though both were written at about the same time in July of last year, and there are some pretty new mentions in Altmetric.

Now, I know that this is an external, commercial service that probably does not feel like it owes any transparency to anyone (which strikes me as ridiculous, even given the pressure to minimize gamification, but anyways), but I'd still like to know if Stack Exchange as a platform has some knowledge of how this works. In particular, given that these guys presumably trawl through the whole corpus of SE posts on a regular basis, I imagine that they've got some special plug they can use, or at the very least they'll have turned up on the server logs. So - is there some method to the madness there? As an extreme, does SE directly report DOI mentions to Altmetric to save them the scraping bit? Or is this the sort of stuff on which SE cannot comment because agreements?

I realize this isn't such an important question, but I'm curious - and ultimately I think it does have some relevance to the impact of posts on this network, particularly as I would argue that mentions in SE posts are on average probably better indicators than social media hits, and it would be nice if they were consistently represented in this sort of quasi-trackback system.

4

As you said - this is external to Stack Exchange, we don't know how they are getting their data.

As far as I know, we don't report DOI mentions to Altmetric (or anyone, for that matter).

It is indeed possible they are scraping - entirely possible they are not doing a good job there.

We don't like people scraping our sites, usually because many do a bad job at it (they don't throttle the scraper and can impact the site - we would block them automatically if too many requests come in in a short time period). This is why we provide an API - this is what they should be using (and may be using - I don't know for sure).

  • OK, fair enough. I would hope that they're on the API and presumably on its logs, but I imagine you don't find it that interesting to trawl through that data to check. – E.P. Feb 13 '17 at 13:07
  • I wouldn't know what to look for, @E.P. - I don't know what IP address they use (and even if I did - perhaps they are doing everything right on our end, but have a bug on theirs?) – Oded Feb 13 '17 at 13:12
  • I would hope for some relatively distinctive user agent or similar self-reporting via whatever mechanism is set in the API requests ('cause that'd be the professional thing to do, right?). If there's no such ID on those logs then yeah, asking you to trawl to try and identify bots by their IPs is too much. – E.P. Feb 13 '17 at 13:22
  • Either way, is this something SE was unaware of? Because I would find that pretty surprising, to be honest. This is an interesting trackback mechanism and potentially a source of high-quality traffic, so (shortcomings and all) it's worth having it well-oiled, I would argue. – E.P. Feb 13 '17 at 13:23
  • We have a lot of API users (and looks like Altmetric are registered users). Given this is an issue with their service - so really should be directed to them. If they do find a problem with our API - then there's something for us to do. – Oded Feb 13 '17 at 13:44
  • Apologies, I absolutely didn't mean to press. I ask because you guys we can talk to, and that's way harder with these publisher conglomerates. Thanks for your time =). – E.P. Feb 13 '17 at 13:47
  • @E.P. - no worries. Just saying - nothing much to do on our end... Altmetric is completely opaque to us too - the data they have is public data and how they use it is up to them... – Oded Feb 13 '17 at 13:51
  • Sure. I just meant that I'd be surprised if you guys had completely no idea that SE data was being used in this half-hearted but positive way (since you're usually very good at keeping on top of those kindsa things). And yeah, the statement 'Altmetric is completely opaque to us too' is pretty much all I wanted to know. – E.P. Feb 13 '17 at 13:53

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .