Introduction

With the rapid development of new media technologies, more people are turning to online sources to seek information about science and scientific developments. Unlike traditional print and broadcast media, new web-based platforms have made it possible for audiences to step out of their passive roles and connect with the communicators themselves [1]. Hence, outside the scientific community, lay audiences are increasingly turning to other sources, besides mainstream journalistic ones, for scientific information [2]. They are also able to directly participate in scientific discussions by commenting on posts on online platforms and even producing their own content. The digital era has also given researchers the opportunity to manage the communication of their research outputs themselves. They can now reach a wider audience online via diverse channels and in different ways, thus a researcher’s scientific outreach is no longer restricted to the traditional academic community. Researchers are bringing attention to their publications on social media platforms like ResearchGate (https://www.researchgate.net) or Twitter (https://twitter.com) in an attempt to gain more readers, citations, and social impact [1].

Altmetrics can be described as new or alternative metrics based on activities on social media for measuring scholarly impact [3, 4]. However, in the literature, the term altmetrics is still very controversial, as it could also be seen as just a new form of informetrics [5], and the distinction between usage metrics and altmetrics is still unclear [6]. Many attempts have been made to define usage statistics. They could be described as an aggregation of the underlying usage data, for example the number of full text downloads or views of a publication [7]. Download statistics can be collected very quickly and provide an immediate measure of how many people have downloaded a publication. However, usage statistics could be a misleading measure of the impact of an article, as they are a simple but inaccurate measure of actual usage [8].

Traditional metrics such as citation counts or the h-index are slow indications of impact, for instance, citations could take years to appear. They also represent a narrow academic audience [9]. In contrast to traditional metrics, altmetrics reflect impact faster, as a tweet could appear within hours of a paper being published. In addition, altmetrics represent both academic and lay audiences [10]. Altmetrics however face some challenges. For example, the heterogeneous nature of altmetric sources [11] is a major challenge, as well as the inconsistency and lack of normalisation of altmetrics [10]. Also, very little is known about the motivation of users and the reasons why they mention and discuss research outputs on social media [3, 12]. Altmetrics are also very susceptible to manipulation [10]. For example, a study shows that automated Twitter bot accounts cause a lot of spam to scientific articles [13]. Download statistics are also vulnerable to manipulation and gaming, although many publishers have built in spam detection to prevent this [8]. A study shows that scholars have engaged in deceptive self-downloads on SSRN to lessen negative social comparisons with their peers [14]. Thus, to ensure data integrity, some strategies have been created, for example, the detection of irregular usage [15] and tracking suspicious activity by PLOS [16], as well as identifying patterns that indicate fraudulent downloads by SSRN [17]. Altmetrics also face many data quality issues [11] and there are no common standards across the different altmetric aggregators [18], though the publisher cooperative Crossref’s DOI Event Tracker pilot [19] may soon result in progress on that front. There have also been efforts made to recommend good practices like from the National Information Standards Organization (NISO) [20] and the European Expert Group on Altmetrics [21].

Since traditional metrics have been criticized for not comprehensively representing scholarly impact, not only researchers but publishers and service providers are starting to consider alternative tools to track the online presence of authors, articles and journals. Altmetric aggregators such as Altmetric.com (https://www.altmetric.com), Plum Analytics (http://plumanalytics.com), and Impactstory (https://impactstory.org) compile a range of altmetrics from different data sources and provide them to researchers to view and track [3, 4]. Some altmetrics are based on events such as views, reads, clicks, downloads, saves, bookmarks, tags, favourites, comments, reviews, links, shares, tweets, likes, or ratings [3]. Some also collect citations [4]. Alternative metrics have been corroborated as useful marketing tools by publishers in promoting articles and journals [22]. For example, Scopus (https://www.scopus.com) introduced “Altmetric for Scopus”, a third party web application that integrates into the sidebar of pages in Scopus, where researchers can track the mentions of different social media metrics and reference managers. Furthermore, editors and authors find it challenging to engage online with different audiences, and many readers find it difficult to locate and give feedback to publications, as these have been shared on different social media sites [23].

The above arguments and challenges motivate us to study the sources (i.e., online channels) that are commonly used to share publications, and to analyse the effectiveness of the sources for sharing a publication (i.e., to determine whether the number of links shared on online channels could be associated with an increase in the number of full text downloads) using Kudos (https://www.growkudos.com) as a case study in this paper. In the next section, we deliberate on the significance of new media in academia and discuss several features and services of social media platforms and altmetric services that provide support for researchers in their efforts to promote their research outputs and increase their outreach. We then look at related studies that investigate these systems. We present our reasons for choosing Kudos as a case study for this paper, we introduce the system and features provided to researchers. We then present the research questions we aim to address and the methodology used for the analysis. After this, the results of the analysis are presented. We then discuss our findings, limitations, and future work.

Related work

The use of social media has been increasing rapidly over the past years. Although its rate of adoption has been slower in academia than in the general population [24], more scholars have started to see its usefulness, and some foresee that it may eventually influence tenure and promotion processes at academic institutions [25]. Scientists have acknowledged professional benefits to using social media tools, for example, a survey of highly cited US nano-scientists claim that the use of social media can contribute to a researcher’s scientific impact [26]. Scholars in the information science and technology field in North America are also starting to use social media tools professionally as they have been found to make research dissemination and connecting with peers more convenient [27].

A comprehensive literature review on the usage of social media in academia and in scholarly communication shows the interest and popularity amongst researchers in using social media platforms [28]. Social media are perceived as effective tools for the discovery and dissemination of research outputs [29]. A study by Nature shows that researchers use platforms like Google Scholar (https://scholar.google.com), LinkedIn (https://www.linkedin.com), ResearchGate, Academia.edu (https://www.academia.edu), Facebook (https://www.facebook.com), Mendeley (https://www.mendeley.com), and Twitter with varying degrees of usage intensity, and for diverse purposes, including maintaining their scholarly profiles, and sharing their research outputs [30].

Specifically, scholars have cited information dissemination as a major benefit of using Twitter [31]. In a study on the usage of Twitter by 587 scientists, it was reported that 32% claim to have posted an estimated number of 1001–5000 scientific tweets (tweets including scientific content or a science-related hash-tag), whereas 22% claim to have posted 101–500 scientific tweets, the majority aiming thereby to connect with fellow scientists [32]. Among digital humanities scholars, studies show that Twitter, functioning both as an information network and a social network, is seen as an important tool for informal communication [33]. Being mentioned on Twitter can help magnify the effect of interactions with journalists and those outside the scientific community [26]. Social impact measures based on tweets, such as the twimpact factor, have been proposed to complement traditional citation metrics [34].

Also, a survey of 2,414 researchers shows that researchers use YouTube (https://www.youtube.com), Slideshare (http://www.slideshare.net), and Flickr (https://www.flickr.com) to disseminate a wider range of research outputs, such as videos, presentation slides, or photos [29]. Meanwhile, in addition to traditional abstracts, some publishers such as PNAS, PLOS Biology, PLOS Genetics, PLOS Neglected Tropical Diseases, Frontiers in Ecology and the Environment, Behavioral Ecology, and Functional Ecology are making summaries available that are aimed at a wider audience [35]. This is seen as a move to make scientific research more visible and transparent. From a researcher's perspective, lay summaries provide them with a way to communicate their research to people outside their specialised field more easily [35]. For example, on Facebook, users are able to provide explanations and accompanying texts when they share links. The comments section also helps to encourage online discussion of an article. Making research more understandable to the general public could help to raise the awareness of the relevance of research outputs, and consequently enable faster adoption of practices [36].

There have been several studies investigating the effectiveness of systems that support researchers in disseminating their works and increasing their outreach. A recent study on Academia.edu attempted to measure the difference in citations between articles posted to Academia.edu and other articles from similar journals, using a sample of 31,216 papers [37]. Results from the study were however controversial as the methodology was found to be flawed and the findings possibly false [38].

In another study, the attributes of philosophy scholars on Academia.edu were examined [39]. Results showed that faculty tend to attract more profile views than students. A gender comparison also revealed that female philosophers did not attract more profile views than males. This suggests that academic capital drives philosophy uses of the site more than friendship and networking. A faculty advantage was also observed in law, history and computer science, except for females in law and females in computer science. Moreover, a correlation analysis found no correlation between traditional bibliometric measures and any Academia.edu metrics for philosophers [39].

In addition to Academia.edu, studies have also been conducted on other platforms such as ResearchGate, and Mendeley. A study on ResearchGate examined if the platform’s usage and publication data broadly reflect existing academic hierarchies and whether individual countries are set to benefit or lose out from the site. Findings indicate a moderate correlation between ResearchGate statistics and other rankings of academic institutions which suggests that ResearchGate use broadly reflects the traditional distribution of academic capital. Country-level analyses also found that Brazil, India, and some other countries seem to have a disproportionately large usage of ResearchGate. On the other hand, academics in China, South Korea, and Russia are making relatively little use of ResearchGate and may be missing opportunities offered by its use [40].

Mendeley readership counts have also been compared with citations for various social sciences and humanities disciplines. Results show that the overall correlation between Mendeley readership counts and citations for the social sciences was higher than for the humanities. There were also low and medium correlations between Mendeley bookmarks and citation counts in all the investigated disciplines, implying that they could be measuring different aspects of research impact [41]. A recent literature review [4] gives an overview of similar studies comparing altmetrics with citation, and a meta-analysis shows overall a low to medium correlation between most altmetrics and citation counts. Also, in several studies, altmetrics were compared with usage metrics and other altmetrics [4]. For example, in [42], a medium to high correlation was measured between ScienceDirect downloads and Mendeley readership counts, and also between ScienceDirect downloads and Scopus citations. Considerable differences were found in download statistics on the levels of discipline, journal, and document types [43]. Also, according to a study comparing downloads to citations in four disciplines (arts and humanities, computer science, economics, econometrics and finance, and oncology), 50–140 downloads could potentially correspond to one citation [44]. However, another study concludes from the log analysis of viewing full text articles, that scholars often download an article first and decide later about its relevance [45]. Also, publishers may not always be able to track all full text downloads of a publication as these might be available for download elsewhere [43]. Thus, it could be said that downloads are a statement of intent to use the publication [44], but this might not necessarily bring about any further use of the publication.

What however has not yet been extensively investigated is whether the outreach efforts of authors, (e.g., sharing their publications via social media and other online channels), could be associated with an increase in a publication’s metrics (e.g., the full text downloads of a publication). A study in the clinical pain sciences, albeit with a small sample of 16 PLOS ONE articles, does show that sharing via Facebook, Twitter, LinkedIn and ResearchBlogging.org is associated with an increase in the number of views and downloads of a publication, although the absolute effect size might not be considered substantial [46]. Thus, this motivates us to investigate how researchers share their work on social media and to try to measure how effective this is by analysing Kudos as a case study in this paper. Kudos is distinct from most of the services mentioned above as it enables researchers to manage multiple communication channels (rather than being an outreach channel in and of itself), and to track communications against a range of metrics (including citations, downloads and altmetrics) to understand their outreach efforts over time. Kudos achieves this by generating “trackable” links that researchers can share on social media as well as via other communications media such as email, or even offline communications (e.g. presentations, reading lists). Services like Twitter, Facebook and LinkedIn enable researchers to promote their research, but results are measured in terms of social media metrics such as views, shares and likes and not in terms of publication metrics such as downloads and citations; scholarly collaboration networks such as ResearchGate and Academia.edu are similar in that they enable promotion of work but with limited metrics, in these cases further restricted to other users within academia. Altmetric services such as Altmetric.com, Impactstory, and Plum Analytics track attention to research outputs across diverse data sources, but don’t show the full range of researchers’ own communication and dissemination efforts. Similarly, publisher websites and institutional repositories host content and provide a range of metrics—including both altmetrics and traditional metrics such as downloads and citations—but do not track researchers’ communications to determine where these may be improving results. Kudos combines data from all these sources, thereby building a dataset containing both communications data and metrics, and is therefore uniquely able to support analysis of the association between outreach efforts and metrics. Thus, due to these unique features and the data Kudos captures and cross-references, we select Kudos as a case study to investigate in this work.

Case study: Kudos

Kudos (https://www.growkudos.com, launched in May 2014) is a web-based service that claims to help researchers, institutions and funders maximise the visibility and outreach of published articles. Kudos enables analysis not only of the current outreach of a publication, but also how that outreach has been achieved.

Kudos encourages researchers to take the following “actions”:

• Share: users can generate a link to their publications and share it online via Social Media, e.g., on Facebook, Twitter or LinkedIn, or they can also share the link through other channels e.g., email.
• Explain: users can provide a plain language explanation describing their work in a simpler manner.
• Enrich: users can upload links to additional resources, e.g., datasets or slides to supplement their publication.

An overview of Kudos is shown in Fig 1, where a publication (essentially its DOI from its publisher) claimed by an author in Kudos can be explained and enriched with a dataset, slides, images or code. The author can also share a link to the publication by generating a share referral and sharing this via social media or other channels. Kudos collects daily full text downloads about this publication from its publisher, as well as the daily Altmetric Score from Altmetric.com.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Overview of Kudos.

https://doi.org/10.1371/journal.pone.0183217.g001

Taking Kudos as a case study, we address the following research questions:

1. RQ1: Which online channels are most commonly used to share a publication on Kudos? Does the career level or discipline of the authors affect the choice of how publications are shared via Kudos?
2. RQ2: Which online channels are most effective for sharing a publication on Kudos? How do the numbers of links shared on online channels via Kudos correlate to an increase in the clicks on this link across different career levels and disciplines of the authors?
3. RQ3: Do any of the actions offered by Kudos have an association with the full text downloads of a publication?

Methodology

In the methodology section, we present the data extraction process and metrics available for the analysis. We then explain the calculation of the increase in metrics which is used in the analysis. In addition, the methods used to recode career levels and disciplines are also presented in this section. We also present the data analysis methods applied to address each of the three research questions RQ1, RQ2, and RQ3 mentioned above.

Metric calculation

For our analysis, a metric increase was defined as an increase in a metric (i.e., full text downloads or share referrals) between the date the publication was first claimed in Kudos and the date of extraction from Kudos (30 January 2016) as shown in Fig 2. A single publication could be claimed by multiple users on different dates in Kudos, thus for such cases, the first available claim date is selected. If there was no change in the metric between the claim date and extraction date, the increase is taken as 0. If the claim date was the same as the extraction date, the increase is also taken as 0.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Metric increase calculation.

https://doi.org/10.1371/journal.pone.0183217.g002

The assumption was made that the metrics are cumulative scores, thus the metric on the extraction date could only be greater or equal to the metric on the claim date. Unfortunately, the Altmetric Score from Altmetric.com does not respect this assumption as sometimes the Altmetric Score experienced a decrease in value due to changes in the calculation of the Altmetric Score by Altmetric.com. Furthermore, Kudos “trackable URLs” were only recognised by Altmetric.com following a change made part way through the time period represented in the dataset for this project. Thus, the Altmetric Score could not be considered further in the analysis.

Results and discussion

We present our findings of the analyses in this section, we first show the results of the descriptive analysis, and then we present the results of each research question.

Descriptive findings

The total number of publications with actions in Kudos was 22,170. Of these, 4,610 (20.8%) had been shared, and 20,775 (93.7%) had been explained. Of those explained, 4,499 (21.7%) had been further enriched. Table 1 gives an overview of the metrics available for the publications having respective actions in Kudos. Of the 4,610 publications shared on Kudos, 9.8% have full text downloads and 32.2% had share referrals. Amongst publications with explain actions (in total 20,775), 22.8% have full text downloads, and 5.7% have share referrals. From the 4,499 publications with enrich actions, 67% have full text downloads, and 4.5% have share referrals.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Publications with actions and their metrics.

https://doi.org/10.1371/journal.pone.0183217.t001

Findings for RQ1

Fig 3 shows the percentage distributions of the different social media channels via which Kudos users shared their publications. Among the social media channels, Facebook is the most popular with 55.3%, followed by Twitter with 42.1%, then LinkedIn with 21.0%, and other channels (this includes sharing via email, other websites, networks, or offline) with 16.7%. However, LinkedIn was introduced by Kudos as an integrated sharing option later than Facebook and Twitter, in July 2015; when we only look at the 2,488 publications that had been shared after 07 July 2015, 62.3% had been shared via Facebook, followed by LinkedIn (38.9%), Twitter (35.2%), and other channels (4.5%). The distribution of shares on social media and other channels before and after 07 July 2015 are shown in S1 and S2 Figs.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Distribution of shares via social media and other channels.

Percentages of publications on Kudos with share actions (n = 4,610) that were shared via the three social media channels: Facebook (2,551), Twitter (1,941), and LinkedIn (969), as well as via other channels (772). Some publications were shared via multiple channels and some on different share dates. Multiple shares on the same channel on the same share date were counted only once in this analysis.

https://doi.org/10.1371/journal.pone.0183217.g003

As shown in Fig 4, among the 4,112 authors who first claimed and shared their publications on different social media channels in Kudos, the majority were faculty (58.9%), followed by researchers (16.8%), professionals (12.3%), and students (6.2%). In terms of discipline, of the 4,012 first authors who claimed and shared on social media channels (shown in Fig 5), 28.6% of authors were from Social Sciences, 26% from Natural Sciences, 18.9% from Medical and Health Sciences, and 14.7% from Engineering and Technology. Only 7.6% of authors from the Humanities, and 1.6% from Agricultural Sciences claimed and shared their publications in Kudos.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Distribution of career levels of authors.

Percentage distributions of self-reported career levels of authors (n = 4,112), who first claimed and shared their publications via Kudos. Of the 4,112 authors, 2,420 were faculty, 689 researchers, 506 professionals, 256 students, and 241 did not specify their career levels or stated “other”.

https://doi.org/10.1371/journal.pone.0183217.g004

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Distribution of disciplines of authors.

Percentage distributions of self-reported disciplines of authors (n = 4,012), who first claimed and shared their publications via Kudos. 1,147 authors were from Social Sciences, 1,045 from Natural Sciences, 759 Medical and Health Sciences, 590 from Engineering and Technology, 304 Humanities, 65 Agricultural Sciences, and 102 from other disciplines.

https://doi.org/10.1371/journal.pone.0183217.g005

Career level differences in sharing on different social media channels.

Fig 6 shows that around 60% of all career levels (professionals, students, researchers and faculty) shared their publications on Facebook, but, a Chi-Square test found no significant difference between the different career levels (p = .832, see S1 Table). Compared to students (38.3%) and faculty (40.8%), professionals (54.0%) and researchers (48.6%) used Twitter more as a sharing platform in Kudos, and this was statistically significant (p < .01, see S2 Table). In terms of sharing on LinkedIn, there was no significant difference among the different career levels (p = .170, see S3 Table). S4 Table shows the findings for “Other channels”.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Distribution of authors, who shared their publications via social media, by career level.

Professionals (n = 506), students (n = 256), researchers (n = 689), faculty (n = 2,420), and other career levels (n = 241) shared their publications via Kudos on Facebook, Twitter, LinkedIn, and other channels. Most authors shared their publications on multiple channels.

https://doi.org/10.1371/journal.pone.0183217.g006

Disciplinary differences in sharing on different social media channels.

As seen in Fig 7, overall, more than 50% of the authors across all disciplines shared their publications on Facebook. In particular, authors from Agricultural Sciences and Humanities shared their publications the most on Facebook (66.2%, and 60.5% respectively). Significant differences among the disciplines were found (p < .01, see S5 Table). The distribution varied for using Twitter as a sharing platform among disciplines, from 36.3% for Engineering and Technology to 50.8% for Agricultural Sciences and 50.7% for Social Sciences. This disciplinary difference is significant at p < .01 (See S6 Table). More than 10% of authors from all disciplines shared their publications via LinkedIn and this was found to be statistically significant (p < .01, see S7 Table), with 25.1% from Social Sciences and 24.7% from Engineering and Technology. S8 Table shows the findings for the group “Other channels”. Interestingly, 34.7% more of authors from Engineering and Technology, and 36.4% more of authors from Humanities shared their publications on Facebook than on Twitter.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. Distribution of authors, who shared their publications via social media, by discipline.

Authors, who shared their publications via Facebook, Twitter, LinkedIn, and other channels were from Social Sciences (1,147), Natural Sciences (1,045), Medical and Health Sciences (759), Engineering and Technology (590), Humanities (304), Agricultural Sciences (65), and other disciplines (102). Most authors shared their publications on multiple channels.

https://doi.org/10.1371/journal.pone.0183217.g007

Findings for RQ2

The findings in Table 2 show that there are small but statistically significant Spearman correlations between the number of links shared on the three social media channels and an increase in share referrals. Sharing on Twitter shows a small correlation (rho = .28, p < .01), followed by LinkedIn (rho = .12, p < .01), and Facebook (rho = .11, p < .01). The results are based on n = 1,485 shared links that had been clicked.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Spearman correlations between shares on social media and an increase in share referrals.

https://doi.org/10.1371/journal.pone.0183217.t002

The correlations between the number of shares on social media channels and an increase in share referrals by career level and discipline are also investigated. As shown in Table 3, on all the three social media channels, small but statistically significant correlations are found between the number of shares made by researchers and an increase in share referrals. For students, results show a medium statistically significant correlation for Facebook (rho = .32, p < .01). For faculty, a small correlation is found for LinkedIn (rho = .12, p < .01), and a medium correlation for Twitter (rho = .32, p < .01). A small correlation is also found between the number of shares made by professionals and an increase in share referrals on Twitter (rho = .27, p < .01).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Spearman correlations between shares on social media and an increase in share referrals by career level.

https://doi.org/10.1371/journal.pone.0183217.t003

Small to large Spearman correlations between the number of shares made on Twitter and an increase in share referrals are seen across all disciplines in Table 4. The correlation results (ranging from small to medium significant correlations for all channels) show that the more links shared by Medical and Health Sciences authors, the more likely they are to be clicked on.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Spearman correlations between shares on social media and an increase in share referrals by discipline.

https://doi.org/10.1371/journal.pone.0183217.t004

Findings for RQ3

Overall for all publications, the Treatment group (n = 4,858) had a significantly higher median average of 149 increase in full text downloads (23.1% more) per publication in comparison to the median average of 121 of the Control group (n = 4,866). The Mann-Whitney U value was found to be significant at (U = 10669164, Z = -8.31, p < .001). The results were not remarkably affected by filtering by document type. When considering only document types: article, article in press, and conference paper, the Treatment group (n = 3,961) still had a significantly higher median average of 158 increase in full text downloads (23.4% more) per publication as compared to the Control group (n = 3,851) with a median average of 128 increase in full text downloads per publication (U = 6929816, Z = -7.00, p < .001). The results are shown in Table 5.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Mann-Whitney U test for different document types.

https://doi.org/10.1371/journal.pone.0183217.t005

For the following analyses, we only consider the document types: article, article in press, and conference paper. We drilled down to investigate an increase in full text downloads for different publication years, different journals, and different disciplines of the first authors who claimed the publications. The results for the different publication years are shown in Table 6. The results show that for almost all years, the Treatment group had higher median averages than the Control group. In particular, this difference is significantly higher in 2014 (U = 252948, Z = -4.64, p < .001), in 2012 (U = 191595, Z = -6.17, p < .001), and in 2011 (U = 76438, Z = -6.21, p < .001). 2016 and the other earlier years (including 2008) had too small sample sizes of n < 40 for the Treatment group (see S3 Fig), thus these comparisons need to be interpreted with caution. Some years could not be considered in the analysis at all, as such, not all years are presented here.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 6. Mann-Whitney U test for different publication years.

https://doi.org/10.1371/journal.pone.0183217.t006

The results for individual journals are shown in Table 7. Journals that had too small sample sizes of n < 40 for the Treatment group, were not considered in the analysis. The Treatment group showed a significant higher increase in full text downloads than the Control group for 7 of the journals: Analyst, Chemical Communications, CrystEngComm, Dalton Transactions, Journal of Materials Chemistry, Physical Chemistry Chemical Physics, and RSC Advances. However, several of the journals presented in Table 7 (including CrystEngComm and Dalton Transactions), have rather small sample sizes for the Treatment group and thus the results should be interpreted with care.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 7. Mann-Whitney U Test for different journals.

https://doi.org/10.1371/journal.pone.0183217.t007

The results of the analysis of the increase in full text downloads by disciplines are presented in Table 8. The Treatment group showed a significant higher increase in full text downloads than the Control group for the disciplines Natural Sciences, Engineering and Technology, and Medical and Health Sciences. The sample sizes for some of the disciplines were rather small (n<40).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 8. Mann-Whitney U test for different disciplines.

https://doi.org/10.1371/journal.pone.0183217.t008

Conclusion

In this paper, we investigate the most effective actions that researchers could take, and channels that they could use to communicate their research outputs, in terms of the increase in full text downloads of the publications. The results show that Facebook is the most popular channel, but that links shared on Twitter are more likely to be clicked on. Moreover, we also find that actions performed in Kudos, such as sharing via social media and other channels, are associated with 23.1% more full text downloads. Citation counts were not analysed in this study so its findings do not claim that an increased rate of full text downloads is associated with a higher citation count in future.

We believe the findings from this study will be useful for individuals and institutions to improve and update understanding of researchers’ outreach efforts and the potential increase in a publication’s metrics. The findings may encourage researchers to use communications media more strategically for building the readership of their work. The findings could also motivate improved provision by publishers and service providers of metadata, metrics and services to support further analysis.

In the future, it would be interesting to gain more understanding of the scholarly community’s strategies of promoting their research, their opinion of disseminating research outputs on social media, and how they keep track of their research outputs. We could also study the information sharing behaviour, habits, and experiences of users on different social media channels, considering their demographics, such as the age group of the authors, career level, discipline, geographical location, etc. Furthermore, sharing via other channels such as wikis, blogs, forums, etc., could also be studied further.

Supporting information

Acknowledgments

The research project “Altmetrics: Rethinking And Exploring New Ways Of Measuring Research” is supported by the National Research Foundation, Prime Minister’s Office, Singapore under its Science of Research, Innovation and Enterprise programme (SRIE Award No. NRF2014-NRF-SRIE001-019).

This research paper is a result of a joint research collaboration between Kudos and CHESS, Nanyang Technological University, Singapore. We wish to express our appreciation to Logan Scott and David Sommer for their support during the project. In particular, to Logan for data curation and data extraction from the Kudos system, and to David for discussions regarding the overall goals of the project.