The Now-Defunct ResearchGate Score and the Extant Research Interest Score: A Continued Debate on Metrics of a Highly Popular Academic Social Networking Site

1 ResearchGate (RG) Metrics: Now-Defunct RG Score and the Extant Research Interest Score (RIS)

Currently, one of the most popular academic social networking sites (ASNSs) is RG.^[1] The popularity of ASNSs stems from academics’ desire to use them to amplify their reputation (Nicholas, Clark, & Herman, 2016) and influence (Desai, Mehta, & Rana, 2024) by making their work – mainly in the form of published research – more visible (D’Alessandro et al., 2020; Francke & Hammarfelt, 2022; Ostermaier-Grabow & Linek, 2019), thus showcasing their productivity (Vinay, Sampath Kumar, & Shiva Kumara, 2020). It also allows researchers to network with industry, although the presence of corporations’ researchers on RG was, until fairly recently, still limited (Yan, Liu, Chen, & Yi, 2020). RG provides a simple yet powerful platform to achieve these objectives, and its metrics serve as altmetrics, or proxies for scientific impact (Hoffmann, Lutz, & Meckel, 2016; Meishar-Tal & Pieterse, 2017), while the presence of RG is strongly associated with productivity on Google Scholar (GS) (Sánchez-Teba, Rodríguez-Fernández, & Gaspar-González, 2021). RG is more popular than another ASNS, Academia.edu (Boudry & Durand-Barthez, 2020). RG serves as a useful complement to institutional repositories (Eva & Wiebe, 2019), or can even substitute them because it provides a publicly visible version of the record unlike some institutional repositories (Borrego, 2017). However, the deletion of an RG account, for example, through moderation caused by a violation of RG user regulations, may result in the loss of an entire repository and the deletion of that researcher’s RG profile (Tsigaris & Teixeira da Silva, 2019). Although RG is proprietary, there are currently no costs involved with establishing an RG account. This, together with its relatively simple and integrated user-friendly interface, makes it a popular ASNS (Kim & Oh, 2021; Kim, 2018), so much so that it now (December 20, 2024) houses, according to RG, 25 million scientists and 160 million publication pages.

Despite these benefits and popularity, some researchers, even established and experienced ones, may exhibit skepticism about the use of ASNSs to self-promote themselves or their work (Greifeneder et al., 2018). Moreover, popularity can breed risks and abuse or misuse by a portion of users and can be used to advertise unscholarly work, papers, or journals (Memon, 2016) or showcase retracted papers in their unretracted state (Teixeira da Silva & Bornemann-Cimenti, 2017), thereby inviting users to cite retracted work and placing the integrity of their own papers’ content at risk. If voluminous, such cases may bring disrepute to the platform and distrust in the science and scientists that are showcased there. RG regularly adjusts the platform’s dynamics to improve structural features, such as changes in rules, resources, or capabilities, or to adjust to users’ and public’s feedback or needs (Huang, Zha, Yan, & Wang, 2019). Collectively, these aspects have made RG an attractive research subject, with volumes of papers related to RG – as indexed on GS – increasing from between 8 and 19 per year in 2008–2013 to 108 in 2017 (Prieto-Gutiérrez, 2019).

Another attractive feature of RG used to be the RG Score, one of the ten reputational mechanisms that RG used to have (Nicholas et al., 2016). However, this altmetric was phased out in August 2022 (ResearchGate, 2022). The RG Score was determined by four factors: publications, questions, answers, and followers (Figure 1a and b). These, when observed exclusively, did not represent the full complement of scientists’ academic achievements and merits, nor did they represent academic failures such as retractions or suspicious publishing activity such as engagement with predatory journals or publishers (Teixeira da Silva & Yamada, 2023). So much so that an RG Score was only able to follow one growth pattern, an increase, and it could never be moderated nor reduced by negative factors. In April 2022, a notice appeared on the “Score” section of RG profiles, indicating that the RG Score would be phased out in July 2022 (Figure 1b). RG’s description of the RG Score changed between 2017 and 2022 (Figure 2a versus Figure 2b). To serve as a historical scientific record for academic posterity, the full statement is indicated in Appendix 1 (Suppl. file) while the full texts of the 2017 and 2022 descriptions are provided in Appendix 2 (Suppl. file). At that time, a query sent to RG regarding how the RG Score was calculated and why there were some discrepancies and accounts with apparent highly inflated RG Scores was met with a deflective response (Appendix 3; Suppl. file).

Figure 1

The RG Score was only visible in “private” mode, i.e., when a scientist logged into their account, and was not public. The visual of the author’s RG Score is displayed from 2017 (a) and 2022 (b). Sources: (a) https://www.researchgate.net/profile/Jaime_Teixeira_Da_Silva/reputation and (b) https://www.researchgate.net/profile/Jaime-Teixeira-Da-Silva/scores. Dates of screenshots: September 20, 2017 (a) and April 13, 2022 (b). Both URLs that were used to support these screenshots are now defunct (i.e., 404 errors), i.e., these URLs no longer exist, and now jump automatically to the author’s RG account top page (October 2023).

Figure 2

Top of the description of RG Score in 2017, 2022, and 2023. 2017 (a) and 2022 (b). The full textual content can be observed in Appendix 2. Sources: (a and c) https://www.researchgate.net/publicprofile.RGScoreFAQ.html; (b) https://explore.researchgate.net/display/support/RG+Score; and (c) https://help.researchgate.net/hc/en-us/articles/14293512753425. The URLs for (a) and (b) are now defunct and jump automatically to the URL in (c). Dates of screenshots: October 6, 2017 (a); April 14, 2022 (b); and October 18, 2023 (c). It is unclear if there were other variations of this text before, in between, or after these dates.

Several criticisms were leveled at the RG Score by academics. First, given its proprietary nature, the equation used to calculate this metric was never openly, transparently, or publicly disclosed, so it could not be replicated because the precise weighting of the four factors was unknown (Kraker & Lex, 2015; Meier & Tunger, 2018; Teixeira da Silva & Yamada, 2023). Another criticism was that the RG Score could be artificially amplified if users asked unrealistic or a disproportionate amount of questions (Cozma & Dimitrova, 2021; Orduna-Malea, Martín-Martín, Thelwall, & Lopez-Cozar, 2017). While questions and answers (Q&As) at RG served to obtain information, recommendations, or feedback (Huang et al., 2019; Jeng, DesAutels, He, & Li, 2017), some were highly subjective (Li, He, & Zhang, 2016) while others were questionable in nature (Figure 3), i.e., redundant questions and/or irrelevant answers (Desai, Mehta, & Rana, 2023), raising concerns about how they could possibly contribute to this supposedly objective altmetric. Another disturbing aspect of RG Q&As is the ease with which they can disappear (see Figure 3c as an example). Since such URLs are typically not captured as mementos at the Internet Archive, such knowledge is irretrievably lost. Conversely, well-structured questions invoked a wealth of informative answers, and such Q&As tended to amplify the RG Score (Li et al., 2023). Thus, pro-active engagement with RG Q&As required an element of positivity or evidence-based negativity, as well as succinct responses (Deng, Tong, Lin, Li, & Liu, 2019). The RG Score could also be abused to artificially amplify institutional rankings (Thelwall & Kousha, 2015; Yu, Wu, Alhalabi, Kao, & Wu, 2016), with institutional participation and thus researcher visibility not accurately reflecting publication output (Lepori, Thelwall, & Hoorani, 2018). Four examples of RG Scores that were assigned to institutions are provided in Figure 4.

Figure 3

Questions and answers purportedly used to contribute towards the RG Score. Despite this, joking questions (a), meaningless answers (b), or “cloned” answers (c) might have been perceived as unfair contributions to a scientist’s RG Score. Disclaimer: the objective of displaying these comments is not to discredit specific academics, but merely to randomly exemplify what is perceived to be redundant and/or problematic Q’s or A’s. Sources: (a) https://www.researchgate.net/post/Did_you_know8; (b) https://www.researchgate.net/post/Interested_in_doing_research_work; and (c) https://www.researchgate.net/search.Search.html?type=question&query=“I discovered a very interesting utopian Russian novel which predicted Putin’s war plan”. Dates of screenshots (a–c): April 18, 2022. The URL that was used to support the screenshot in C is now defunct (i.e., 404 errors) (October 18, 2023).

Figure 4

The RG score was not only assigned to individual academics but also to research institutes, as a cumulative score of its individual researchers. Examples are provided: Karolinska Institutet, Sweden (a); Harvard University, USA (b); The University of Tokyo, Japan (c); and Curtin University, Australia (d). Sources: https://www.researchgate.net/institution/Karolinska_Institutet (a); https://www.researchgate.net/institution/Harvard_University (b); https://www.researchgate.net/institution/The_University_of_Tokyo (c); and https://www.researchgate.net/institution/Curtin_University2 (d). Date of screenshots: September 27, 2017. The four URLs are basically unchanged between 2017 and 2023.

RG thus serves as a way to amplify the academic profile of a scientist via citations (Bornmann, 2016), number of reads (NoRs), and influence, previously assessed by the RG Score, and now assessed by the RIS. Some claimed that the influence of RG’s citation or literature coverage was superior to that of Elsevier’s Scopus and Clarivate’s Web of Science but less than that of GS (Thelwall & Kousha, 2017a, b). The RIS is not publicly visible^[2] on any RG user’s account, and one must first login to observe one’s own RIS and/or the RIS of other RG users. The “Stats” menu of an RG user’s profile indicates that four factors contribute to the RIS: citations, recommendations, full-text reads, and other reads (Figure 5a). Pressing “View details” leads to a pop-up screen that reveals more details (Figure 5b; Appendix 4; Suppl. file). It is fascinating to note how Q&As, which formed an integral part of the RG Score, are not factored into how RIS is calculated.^[3] The RIS was criticized as merely recycling and collating already existing metrics into a new branded metric, thus adding no new metrics value (Copiello, 2019, 2020; Zhang & Kumaran, 2023). Knudson (2023) exemplified this with a small sample of 60 sports biomechanics researchers who were ranked top according to their GS profiles, noting a strong correlation between RIS and GS and RG and Scopus Hirsch (h)-indexes.

Figure 5

The RIS is not visible to the public and is only visible to users after logging in, in the “Stats” section of their profiles. Top page (a) and “Stats” page (b). In the RIS menu of the “Stats” page, clicking the “View details” button (red box) (b) provides additional details and an initial explanation of how the RIS is calculated, noting specifically that self-citations are excluded from the equation (c). More details about how the RIS is calculated are provided when pressing the “Learn more about the Research Interest Score” button (turquoise blue box) (c). Sources: https://www.researchgate.net/profile/Jaime-Teixeira-Da-Silva/stats (a and b) and https://help.researchgate.net/hc/en-us/articles/14293473316753 (c). Date of screenshots: October 21, 2023. A more recent (September 8, 2024) examination of the explanation of how the RIS is calculated indicated no changes between 2023 and 2024.

Two additional criticisms about RIS, not yet indicated in previous literature, are noted in this article: (1) self-citations are automatically discounted as being invalid for the total citations count, even if they offer valid support to statements, and (2) RIS (like the RG Score) can only increase, and there is no mechanism, e.g., as a negative weighting,^[4] that considers negative social media coverage or negative curricular aspects, such as expressions of concern or retractions. To its credit, RG does explain in more transparent detail (relative to the RG Score previously) how RIS is calculated, i.e., the way in which the four main factors are weighted (see Appendix 4 text). However, like the RG Score, the precise equation is not provided, although it was shown to be strongly correlated with GS, Web of Science, and then Scopus citations, in this order, but not their h-indexes (Memisevic, 2022). Even so, no rational or logical explanation is provided by RG as to why, for example, a citation should have a weighting that is three-fold larger than full-text reads (0.5 versus 0.15, respectively).

Does a shift in focus from a scientific influence (metrics only) to a social altmetric (RG Score and RIS) reflect purely scientific merit (Bornmann & Haunschild, 2017), and is emphasis on engagement rather than on scientific output (i.e., publications) (Copiello & Bonifaci, 2018) a way to amplify self-branding? For example, the lack of regulation of RG Projects allowed the authors of some projects to engage in potentially commercial promotional activity, thereby gaining additional reads and artificially boosting their RG Scores (Teixeira da Silva, 2017).

Even though the volume of full-text reads can be amplified by the “open” (green or gold open access) nature of published papers (Copiello & Bonifaci, 2019; Sababi et al., 2017), confusion by authors regarding the licensing and/or copyrighted nature of their papers, or their papers’ versions of the record, in violation of copyright contracts, resulted in the exposure of a body of copyrighted work in an “open” format. This led not only to legal threats against RG by select publishers (Kwon, 2022) but also to its users who violated publishers’ copyright policies (Jamali, 2017), in part amplified by RG emails encouraging linking and sharing (Cress, 2021). In response to that episode, some publishers adjusted their collaborative relationship with RG to accommodate researchers’ needs in terms of knowledge exchange, as a way to try and defend their copyright without having to revert to legal action (Manley, 2019).

Finally, even though the RG Score ultimately provided an RG h-index, it was always inferior to the GS h-index due to the number of publications that GS considered for indexing relative to RG (Singh, Srichandan, & Lathabai, 2022). Although RG Score was recommended by some for amplifying networks or collaborations (Joshi et al., 2019), or for formal academic institutional use (Yan & Zhang, 2018), due to the complex set of concerns and negative debate related to its validity as a scientific metric, others discouraged its use (Bangani & Onyancha, 2021; Clavier et al., 2021; Orduna-Malea et al., 2017). Conversely, one study found that researchers at highly ranked universities used RG and other ASNSs more than researchers at lower-ranked universities, independent of the discipline (Yan, Zhang, Hu, & Kudva, 2021).

RG users should also be aware that metadata of their papers that were automatically transferred by publishers to RG, for example, of journals that form part of the Springer Nature-RG collaboration, cannot always be manually updated and edited, i.e., the initially transferred metadata of the “in press” version cannot be edited by RG users to update volume, issue and page numbers, or publication date, leaving those papers’ RG entries outdated, incomplete and inaccurate, and only a manual paper-by-paper request to RG to update the records can be considered (Teixeira da Silva, 2023a).

2 How are Reads Factored into RG’s RIS?

A core concept at the heart of using ASNSs is to self-promote and thereby self-preserve one’s legend and publishing (and research) repertoire (Hailu & Wu, 2021). At what level of citations or altmetrics can a scientist claim to have reached “fame” (Sternberg, 2016)? The altmetric score of a paper can serve as a reward proxy to traditional citation-based recognition, at three levels: research impact (Barnes, 2015), societal impact (Holmberg, Bowman, Bowman, Didegah, & Kortelainen, 2019), and knowledge democratization (Daraio, 2021). For that reason, NoRs are in themselves a distinct metric because they can, through readership, amplify general knowledge of a paper (and thus indirectly of an author and even a journal) and lead to more citations (Wasike, 2021), although they might not necessarily result in the citation of that paper, i.e., increased viewership or readership does not always imply increased citations (Banshal, Singh, & Muhuri, 2021). NoRs that formed part of the RG Score, and that still form part of RIS (Figure 5b and c; Appendix 4; Suppl. file), are a subset of this metric that can be used to evaluate or rate scientists’ performance and output (Panda & Kaur, 2023), so there is interest in appreciating cases of abuse that attempt to amplify NoRs and readership (Kirilova & Zoepfl, 2025). It has been argued that preprints (or post-prints) can serve as metrics amplifiers, given that they represent a “duplicated” (but not deemed as unethical) copy of a work, and thus an extended readership network, and could thus be interpreted by some as an unfair metrics advantage (Teixeira da Silva, 2023b).

Appendix 4 text (Suppl. file) informs us that “self-reads” do not count toward reads, following a similar logic for self-citations, although in this case, this rule makes sense, otherwise authors might be constantly clicking on their own papers to artificially boost reads. In addition, reads by the general public are not considered, i.e., one must be logged into an RG account in order for a read to count to another RG account user’s RIS. From that text, a link is provided: “Learn more about how we count reads.”^[5] On that page (see Appendix 5 in the Suppl. file for the full text), RG provides an explanation about reads, their importance, how they are calculated, and who the readers are. Although Q&As are themselves not weighted, reads of Q&As are factored into reads (i.e., NoRs), and thus RIS.

3 Conclusion, Future Outlook, and Limitations

This article takes a historical look at some aspects of the now-defunct RG Score, and why RG might have ceased that altmetric, sourcing debate and criticisms from the published literature. Another RG altmetric, RIS, continues to function and is still displayed on all RG users’ accounts in 2024. While there are positive aspects and more transparency about RIS relative to the RG Score, there are still some worrisome aspects, including the risks of abuse, for example, for excessive self-branding, or the unfair (and potentially unethical) inflation of RG’s metrics and altmetrics, including RIS, through inter-researcher “citation rings” (Kirilova & Zoepfl, 2025). According to RG’s CEO, Ijad Madisch, RG serves to self-brand reputation (Winter, 2015). The use of ASNSs for self-branding and self-promotion (Francke, 2019) requires institutional education regarding how to better employ such altmetrics to promote research units at universities (Wiechetek & Pastuszak, 2022), or their use, including RIS, to rank or otherwise characterize universities (Haris, Ali, & Vaidya, 2023). Almost a decade ago, in a social experiment, a fake researcher profile was created at RG to appreciate how it would continue to gather metrics (reads, etc.) despite being dormant (Murray, 2014). That study should have served as a red flag, alerting RG and the academic community that ASNSs like RG could be used to create fake user accounts and thus create fake, cheap-fake, or deep-fake “scientists,” “researchers,” or aspects of RG accounts (Teixeira da Silva, 2023c), or to “hijack” the accounts of existing researchers who do not have an RG account. Some fake accounts were discovered by Eva and Wiebe (2019). It would thus be important, for those with skills to appreciate valid from fake, including cheap-fakes or deep-fakes, to appreciate the extent of this phenomenon at RG.