Otago's Network for Engagement and Research: Mapping Academic Expertise and Connections
presentationposted on 2020-03-10, 03:52 authored by Sander Zwanenburg
Academic expertise is complex, dynamic, and often encoded in jargon (Auriol et al. 2013). Academics move across institutions and increasingly change the topics of their research (Zeng et al. 2019) They apply their expertise in writing that is often specific to a specialised community.
This makes academic expertise hard to find and to understand. For example, it is difficult for prospective postgraduate students of the University of Otago to find the right supervisor. Likewise, organisations that require expertise for their R&D may not be able to identify available experts. Even within universities, their schools and departments, an understanding of expertise and its applications in collaboration is very limited, complicating the management of expertise and the facilitation of its application.
Currently, to find or understand expertise, one might rely on social networks or digital facilities such as search engines, websites, and academic databases. All of these carry important shortcomings in the search for experts. For example, asking people in one’s social network can be time-consuming and ineffective since individuals’ awareness of expertise in their network quickly fades or becomes outdated when going beyond an intimate inner ring of contacts (Hill and Dunbar 2003). Search engines are optimized for finding relevant pages and documents, not experts (Dudek et al. 2007).
How can we map academic expertise, to make it easier to find and understand?
Our answer is a local and practical one. NEAR, the Network for Engagement And Research is an information system under development, that aims to help its users find and understand academic expertise in the University of Otago. Its proof of concept has been developed in the Otago Business School, with data from and about academic staff. The vision is to build a data warehouse around academics’ expertise and its social context, and to communicate that data visually and interactively through a web application. This can be rolled out to all other schools and divisions in the university.
Essentially, NEAR collects data, integrates and interprets it, and communicates this information to its users. The data collected is a combination of user-inputted data and existing data from other systems. Initially, NEAR only collected basic phonebook-type data from Active Directory, another institutional system, and asked academics to put in detail around their Fields of Research, research methods, Sustainable Development Goals, collaborations, and the Fields of Research of their collaborations. The data input from academic colleagues came with challenges. Initially this relied on an online survey that quickly became so complicated that Qualtrics, the survey provider, had to change its technical specifications. We later developed a custom-made profile system, based on a LAMP stack, where people could log in, and fill out their details. One difficulty that remained was that this required a push, not just one-off but continued over time, to get this data actually collected. The data collection emphasis shifted to other systems that contained data on people’s expertise, and that was maintained and updated elsewhere: the Research Output Database, a Research Management Information System, the Media Expertise Database, but also external databases like Elsevier’s Scopus and Clarivate’s Web of Science.
This shift meant that we started the development of data harvesting and integration protocols. These were all developed in-house in R. They consisted of working with APIs and the resulting output. A current challenge is to infer expertise based on the available evidence. This evidence is based on data on different types of publications, grants, and selfreports and are linked to different classifications of research fields. These will overlap to different extents and it is possible that not all fields of expertise are homogenously reflected in such evidence. Possibly, semantic fingerprints can be applied to enhance accuracy and reduce reliance on particular classifications.
We have communicated our information about expertise and their social context through an interactive web visualisation, as shown in the figure below. The visualisation is written with an R package called visNetwork, which is then embedded in a Shiny app. It consists of a network graph, where each node represents a staff member (colour coded for department) or an external party (red), and each link represents an active collaboration. One can zoom in, and hover over these elements to view their details. They are also searchable by name, department, field of research (and their corresponding disciplines and sub-disciplines), research method, and sustainable development goal through drop-down lists. This highlights those applicable nodes and edges. Highlighted staff members can be emailed with the click of a button, allowing to easily bring together people with like-minded research expertise or interests.
In the next stage of the project, we will develop further the data integration schemes, enhance our algorithm to infer expertise based on this data, and update the interactive visualisation to reflect these inferences. This visualisation should not only help users find fitting experts, but gain an understanding of how these experts sit in a dynamic, social context. For example, given a field of expertise, do the experts form a close-knit group or are they scattered around the university? Deeper insights like these can allow for potent outcomes, such as an email to a strategically positioned expert.
We believe that our approach has the potential to augment popular search engines in an important yet local way. Current search engines are optimized for web pages and online documents (e.g. Google), scholarly output (Google Scholar, Web of Science, Scopus), geographic information (e.g. Google Maps, Yelp). NEAR can offer deeper insights about expertise of individuals by combining institutional and public data. It has the potential to allow its users not only to find the most fitting expert, but also to understand the structure and dynamics of particular areas of expertise. Hopefully, in the future, this will help bridge the demand and supply of expertise, and identify opportunities to leverage more fully what people have developed over many years.
I thank Brian Spisak for his fellow leadership in this project and Caitlin Owen and Lahiru Ariyasinghe for their development support. Further, there are many internal organisations that have contributed to the initiative, including the Otago Business School, the Research Support Unit of the Library, Information Technology Services, and Research & Enterprise. Thank you.
Auriol, L., Misu, M., and Freeman, R. A. 2013. "Careers of Doctorate Holders,").
Dudek, D., Mastora, A., and Landoni, M. 2007. "Is Google the Answer? A Study into Usability of Search Engines," Library Review (56:3), pp. 224-233.
Hill, R. A., and Dunbar, R. I. 2003. "Social Network Size in Humans," Human nature (14:1), pp. 53-72.
Zeng, A., Shen, Z., Zhou, J., Fan, Y., Di, Z., Wang, Y., Stanley, H. E., and Havlin, S. 2019. "Increasing Trend of Scientists to Switch between Topics," Nature communications (10:1), pp. 1-11.
ABOUT THE AUTHOR
Sander Zwanenburg is a Lecturer within the Department of Information Science. He obtained Bachelor and Master of Science degrees from the University of Groningen, The Netherlands, and a PhD degree in Management Information Systems from The University of Hong Kong. Sander’s research interests lies in the fields of the psychology of IT use, the development of metrics, and networks of knowledge. He has published in various Information Systems venues such as the Australasian Journal of Information Systems, Communications of the Association for Information Systems, and the proceedings of the International Conference on Information Systems.