Huiqin Gao | UX Designer
Designer, advocate, supportive leader
Mendeley Copy 2@1x.png

Seeking information with less effort

Research and analysis of publication sharing on Mendeley

Social Information Navigation

Research and analysis of publication sharing on Mendeley

Picking berries from Mendeley’s public groups, by Huiqin Gao


For: Master Thesis

Challenge: How social navigation influence information seeking

Length: 600 hours

Skills: Information science experiment design, data structure, qualitative analysis, social network analysis, rapid prototyping

My role: Lead researcher


Motivated to understand the connection between online social groups, information seeking behavior, and resource management, I investigated the readership of scientific papers shared within and across Mendeley’s public groups (SiG papers)  in comparison with those that were uploaded to the site but not shared (non-SiG papers) for my Master’s thesis.





Information foraging theory

From a literature review I conducted on information foraging, I knew that electronic information seeking behavior is akin to picking berries in a forest. Users often begin the process uncertain about the information available to them and uncertain how to search for said information.

Bates’ Berrypicking Information Retrieval Model (source)



I hypothesized that Mendeley’s public groups were efficient starting points for users seeking scientific papers because the sharing of information in a social context naturally filters for value and relevance.

Data collection

I started by conducting quantitative research on the readership of SiG and non-SiG papers to ascertain the value of the documents. To do this, I recruited a programmer from Wuhan University School of Computer Science to build a Python crawler to collect data from Mendeley, and I assembled a group of five undergraduate research assistants to manually download data from the Web of Science.

To mitigate bias, I compiled comprehensive datasets:

  • All public groups as of July 22, 2014 (TOTAL: 106,156)

  • All papers shared to one or more public groups (TOTAL: 5,034,736)

As a control, I compiled a representative dataset of non-SiG papers the Web of Science:

  • All publications corresponding to Mendeley’s 25 disciplines and 448 sub-disciplines (TOTAL: 1,854,115)

Once the datasets were compiled, I investigated the relationships between Mendeley groups through their co-sharing of papers.




Public groups have more valuable papers


“Picking berries? Why not check out the bowls first!”

Because the data sets were verified as highly-skewed, I ran a Spearman correlation to test whether readerships are reliable index of papers’ values. Results indicated a positive answer.

Next, I conducted a Mann-Whitney U test to analyze the difference between SiG paper and non-SiG papers’ readership levels. This led to my main finding:

SiG papers are significantly more valuable than non-SiG papers.

Therefore, I was confident to draw my final conclusions: the public groups on Mendeley are valuable proxies for information seeking, because they act as manual filters of related publications.



Internal connection between groups

Because the groups are related to each other by the common publications they shared, I conducted a social network analysis based on the co-sharing of papers by all public groups to picture out these internal connections.

The analysis revealed how groups are related to each other:

component size 1.png


Redesign & Recommendation

Taking advantage of my findings on social network analysis, I redesigned the group search page to help users find information more efficiently.

A rapid prototype design for Mendeley’s group search results

A rapid prototype design for Mendeley’s group search results


I added a sorting switch so users could rank the search result using number of papers (wealth of documents), number of members (group popularity), or average readership (value of groups’ documents) of all the papers shared in the groups.

I also suggested that the findings on the interrelation of public groups could be used to improve future information seeking algorithms.




The research was published and presented at iConference 2015 and 2016:

Gao, H., Hu, C., & Jiang, T. (2015, March). An exploratory study of paper sharing in Mendeley's public groups. Paper presented at iConference 2015.

Jiang, T., Gao, H. (2016, March). Are Mendeley's public groups effective aggregators of high-value papers? An analysis based on paper readerships. Paper presented at iConference 2016.