ResearchGate, heal thyself… please? (UPDATED)

People reading this blog (I sure hope someone reads it *bites nails*) may be familiar with the name ResearchGate. It was envisaged as a social networking site focused on scientists; founded in 2008 by Ijad Madisch and Sören Hofmayer, both physicians, and Horst Fickenscher, a computer scientist, the site’s stated mission statement is: to connect researchers and make it easy for them to share and access scientific output, knowledge, and expertise. Although not unique (or the only player) in this field, ResearchGate offers several features which are valuable for academic collaborations:

• Making connections, the first step to collaboration: The site’s algorithm works behind the scenes to match users to colleagues in the same institution, scientific peers, co-authors from different institutions, as well as specialists in a given field. Being able to connect with co-authors conveniently is a useful feature, and ResearchGate makes it easy by pulling out the names of co-authors from the user’s publication list.
• Asking questions: The site allows research professionals to ask questions about research-related problems, and the questions quickly reach users spread throughout the world. If, in course of an experiment, a researcher comes across a methodological problem, chances are that similar problems may have been already encountered (and perhaps, solved) elsewhere by someone else, who is, therefore, in a position to offer a helpful word or two. This is a feature that I have used successfully.
• Sharing publications: In the user profile, the user has the ability to upload their publication list, and under certain circumstances*, their papers, which facilitates the sharing of academic publications. It is also possible to request authors for their publications, and send papers to requesters, via messaging. ResearchGate also allows the archival of conference presentations apart from the traditional journal format.
• Getting statistics: This is an interesting and useful feature, in which ResearchGate works in the background to track views and downloads, as well as citations of papers in the user’s publication list, and offers the option of sending a weekly summary of such academic access via email to the user.

[*NOTE: The sharing of published papers is in a grey zone. ResearchGate encourages sharing of full texts via the process known as ‘self-archiving’, which means that an author uploads an electronic, full-text version of their own paper to ResearchGate’s servers. This, understandably, has the potential to enhance the visibility and readership – and therefore, the academic impact – of a scholarly work. However, the current reality is that most publishers of such work like to retain a strict control over its distribution by withholding self-archival permissions and claiming copyright violations. ResearchGate helpfully provides a color coding scheme which indicates whether the publication can be freely uploaded (generally Open Access publications, as in the PLoS journals) or not.]

In addition, the site also maintains a research-oriented job board, which automatically matches open jobs to the user’s skill sets. As a concept, therefore, ResearchGate is excellent and potentially immensely useful. And the beauty of it is that most of the hard work is done in the background by an algorithm which looks up and matches publications to authors and citations, and so forth, with zero user input.

Which brings me to why I have lately been rather disappointed with ResearchGate’s algorithm. Let me explain.

I was alerted to a weird situation by a scientist friend of mine, a former lab-mate with whom I have co-authored several research papers. He indicated that at ResearchGate, he was seeing strange names coming up as his “coauthors”, and upon digging around, he had discovered that someone unknown to either of us was claiming co-authorship of papers which we had authored. Intrigued, I went in to check, and figured out what the problem was.

We had a third friend/lab-mate as the co-author. Let’s assume this friend’s name to be Dr. An Excellent Researcher, PhD. The papers we coauthored naturally list her with the usual abbreviations, such as ‘An E. Researcher’, ‘An Researcher’, or ‘A E Researcher’. The person claiming authorship of our papers was named Dr. Other Researcher, and all his collaborators were showing up as suggestions for my friend’s and my “coauthors”. I wrote to this Dr. Researcher, apprising him of the situation, and requesting him to remove from his ResearchGate profile all the publications belonging to our friend, Dr. A E Researcher. Dr. O Researcher was rather nice, and he promptly complied.

I had absolutely no reason to believe that this apparent “takeover” of publication authorship was deliberate or malicious in any way. I chalked it up to the fact that the common last name, “Researcher”, fooled the ResearchGate algorithm into adding A E Researcher’s papers into O Researcher’s profile. That this was indeed the case became even more apparent when I looked into the suggestions made by the ResearchGate algorithm to include papers authored by me (click the screenshot to embiggen):

As I went on clicking “This is not me”, it was clear that the ResearchGate algorithm was offering me tens of publications by unknown researchers – A K Datta, K K Datta, and other assorted folks with ‘K’ and ‘Datta’ anywhere in their names – for inclusion under my name in my profile. (NOTE the URL on top, which would allow a user to edit their automated publication matches.)

To my mind, this is a MAJOR FAIL of the citation searching algorithm at ResearchGate. And the funny (funny = peculiar, not funny = ha ha) thing is that this was not the first time I declined authorship of papers by these researchers at ResearchGate by clicking ‘this is not me’. Somehow, the fact that I had declined authorship of papers not authored by me appeared not to stick in the ResearchGate system, which – again – showed me the same papers from same authors just now. This morning was a slow morning for me because a meeting got canceled, but who amongst us really has the time to go on weeding out hundreds of publication suggestions from our profiles? The problem is that if we are not proactive in this, O Researcher takes credit – even unwittingly – for scientific work done by A E Researcher.

And what transpired this morning was beyond hilarious. I received an automated email informing me that the author of an article I cited was ‘waiting for my feedback’ – which was mightily surprising, given that the article in question was published in… 1992! I clicked on the link which would take me to the publication, and reached the page of the “author” who was ‘waiting for my feedback’. And lo and behold! Another gaffe by the ResearchGate citation matching algorithm.

Observe how the algorithm erroneously matched this researcher, Jhilam Mukherjee, with the article author Jean Mukherjee simply on the basis of the first initial and last name; not only that, it mangled the name of the second author, Dr. Matthew D. Scharff, a much-respected researcher in the field of immunology. Even weirder, a quick click on the researcher’s profile link informs me that this person works in the field of Information Technology, and yet, the 32 publications featured by ResearchGate in this person’s profile were almost entirely made out of Jean Mukherjee’s publications, which are in the field of microbiology and immunology; I should know, because many of those are cited in my PhD thesis and other publications. ResearchGate helpfully tells me that I “cited Jhilam in a publication”. Hell no, I didn’t; nah-uh.

ResearchGate certainly doesn’t make it easy to contact them with site-specific questions or concerns such as these, what with the “Contact Us” URL being placed only at the very bottom of the pages in small, grey text – as you can see from the first screenshot above. In addition, there appears to be no easy way to apprise the ResearchGate team of such mistakes. This morning I wrote to ResearchGate using their contact form, indicating that the increasing frequency of these errors is alarming. I am yet to hear back from them. We shall see.

UPDATE (September 23, 2015): ResearchGate community support has responded via email to my message to them.

Thanks for getting in touch. Our publication search algorithm is designed to help you quickly confirm authorship of your publications on ResearchGate. The ease of this one-click process means mistakes can sometimes be made, so if another researcher has claimed to be your co-author, it is usually the result of an honest mistake.

If you notice that someone has mistakenly claimed co-authorship of your publication, please provide us with details and a link to your publication’s ResearchGate page.

We take erroneous authorship claims very seriously and review each case manually, so please bear with us.

I think the emailed response, sadly, falls short of addressing the chief complaints about the system that I have outlined above, and there is no guarantee that the citation search algorithm will be improved ever. This post has been retweeted several times, @-tagging the ResearchGate Twitter handle, but I haven’t had anyone reach out to address these issues.

UPDATE 2 (October 14, 2015): 1. I am still receiving “matches” of “my publications” from ResearchGate, which are in reality FAR away from mine. 2. “Jhilam Mukherjee” still appears in my “Related Researcher” list saying I “cited their publication”.