Introducing Network Analysis in the Classroom

In March and May 2024, as a Training Fellow of the Centre of Data, Culture, and Society (CDCS) at the University of Edinburgh, I had the opportunity to teach two courses on network analysis. The first course is an introduction to performing network analysis with Gephi, while the second focuses on applying Gephi to network analysis beyond human entities, such as ownership and financial networks. This blog post illustrates the challenges faced by a history researcher in introducing network analysis (or more generally digital humanities) to audiences with diverse disciplinary backgrounds, and in using online, publicly available resources in preparing teaching materials. The courses also serve as a platform for participants to experiment with network analysis in their own discipline and eventually apply it to their projects.     

The Challenges of Teaching Network Analysis Through Historical Examples

The two network analysis courses were open to students and academics from a wide range of research stages and disciplinary backgrounds at the University of Edinburgh. As a researcher in modern East Asian history, my biggest challenge was selecting historical or contemporary examples to create datasets for hands-on experience during the courses. Since most participants did not have a background in history and cannot read Chinese characters or kanji, it was expected that they would be confused by a historical network dataset containing hundreds or thousands of romanised Chinese and Japanese names — especially if the dataset contained names in different romanisation systems, such as Wade-Giles and Pinyin. Even if they could overcome the challenge of reading Asian names, without previous knowledge of Asian history they would have difficulty in interpreting the historical networks. However, being able to read networks is essential to understanding some basic concepts of network analysis, such as modularity and centrality measures. In other words, I needed to create historical networks and datasets that were accessible to non-experts in East Asian history in order to introduce network analysis to a wider audience.

Therefore, I used the networks of the Ho Tung family 何東家族 as one of the examples to introduce the use of Gephi. The family is centred around Sir Robert Ho Tung 何東, his brother Ho Fook 何福, and their half-brother Ho Kom-tong 何甘棠. Robert Ho Tung and Ho Fook are Eurasian (their father is Dutch and their mother Chinese), while Ho Kom-tong is Chinese. The Ho brothers became compradores for British companies and were later appointed legislative councillors by the colonial government. They also established maternal ties with other Eurasian families in Hong Kong, such as the Lo’s 羅文錦家族, the Zimmern’s 施炳光家族, and the Wong Kam-fuk’s 黃金福家族. Recent research has used network analysis tools such as Gephi and Rhumbl to explore social networks in nineteenth-century Hong Kong, which was featured in a Digital Orientalist blog post. I used the networks of the Ho Tung family as an example for two reasons: 1) As a Eurasian family, many of its members have English first names, and some had surnames in English or other European languages, making the Ho family networks more accessible to non-experts in East Asian history; 2) the networks contained 111 figures, which is a manageable size for beginners. I have also included a brief introduction to the Ho family in the handout, which will be useful in analysing the networks. 

Some might ask why we need to use Gephi or other software to map social or family networks instead of drawing them by hand, as historians, sociologists, and (investigative) journalists have been doing for several decades. We may have similar questions in mind. Of course, it is possible to draw networks manually, so why are network analysis tools more useful for researchers? What additional information or statistics can researchers gain by using digital tools? I decided to answer this question together with the participants. First, I give them three spreadsheets, which are the lists of the members of Robert Ho Tung’s family, the Ho Fook’s, and the Ho Kom-tong’s. The lists also included relationships between family members. Figure 1 is the spreadsheet of Robert Ho Tung’s family, the second column shows that Ho Tung is the brother of Ho Fook, while Ho Tung is the half brother of Ho Sui-ting 何瑞婷 and Ho Kom-tong. The first activity in class is to draw the network of each family by hand. This is not very difficult for the participants as each family contains only about 10 to 30 entities (nodes) and relations (edges). The second activity is to draw by hand a network that includes the families of Ho Tung, Ho Fook, and Ho Kom-tong, and to identify the central figure(s), connectors, bridges, and gatekeepers in the network. This task is challenging because it is difficult to manually draw a network with more than a hundred nodes and edges on paper, and also to prove who are the central figures, connectors, etc. The second activity shows the advantages of utilising digital tools in network analysis, namely: To efficiently map an immense and complex network, and to quantitatively identify the important nodes in the network through the statistics obtained from centrality measures (figures 2 and 3).

Figure 1: Spreadsheet of Robert Ho Tung’s Family

Figure 2: The Network of the Ho Tung Family

Figure 3: Centrality Measures of the Ho’s Family Network

The Use of Online and Publicly Available Resources in Network Analysis

There are plenty of online, publicly available resources that can be used to prepare teaching materials for network analysis. Using data from Harvard University’s China Biographical Database Project (CBDB), a ‘relational database’ geared towards social network analysis and prosopography, I demonstrated two of Gephi’s features — GeoLayout and ExportToEarth — to the class using the social networks of key leaders of the Xiang Army 湘軍 coalition. These networks highlight the importance of local (kinship) ties in coalition building. 

Figure 4: Overlay of Zeng Guofan’s 曾國藩 Kinship and Friendship Networks on Google Earth

I also employed the data from the Webb-site to demonstrate the use of Gephi to reconstruct financial networks. Webb-site is the website of David Webb, a retired investment banker and activist investor. On 17 May 2017, he published an article entitled ‘The Enigma Network: 50 stocks not to own’, in which he argued that investors should not invest in 50 Hong Kong-listed companies because these companies had cross-owned each other’s shares. This allowed the ‘big bosses’ behind the ‘Enigma Network’ to easily manipulate the prices of these stocks and therefore damage investors’ interests. Webb’s article had shocked the market and the Securities and Futures Commission of Hong Kong began to investigate his claims. As a result, several executives of the companies in Webb’s networks were arrested and prosecuted, although most were eventually acquitted.

The website provided comprehensive data on the ‘Enigma Network’, including the parent and subsidiary companies of the 50 companies in the network, as well as daily shareholding information from the Central Clearing and Settlement System (CCASS), a computerised book-entry clearing and settlement system for transactions in securities listed on the Stock Exchange of Hong Kong (HKEX). The CCASS data allows us to identify the top 10 shareholders of these 50 companies on 16 May 2017, the day before the implosion of the ‘Enigma Network’. (Figure 5). After entering the CCASS data into a spreadsheet and quantifying the weight of the edges in the network according to the proportion of shares a shareholder held in a company, it can then be imported into Gephi to map the network and for further analysis.     

Figure 5: The CCASS Holdings Record of One of the 50 Companies on 16 May 2017

The centrality measures of the network show that the largest shareholders include securities firms, (investment) banks, and some of the 50 companies in the ‘Enigma Network’ (figure 6). I am not suggesting that all these financial institutions, especially the giant banks, are part of the ‘Enigma Network’. They may hold shares in these companies for various reasons, such as profits from short-term investments. 

Figure 6: Centrality Measures of the ‘Enigma Network’

The experiment of using data from the Webb-site and CCASS to reconstruct the ‘Enigma Network’ illustrates the potential use of Gephi and other network analysis software to analyse historical and contemporary investment, financial, and ownership networks from around the world, as well as other types of networks beyond human entities. 

Moving Forward

In addition to playing with different (historical) networks, participants in both courses are encouraged to develop their own projects and experiment with network analysis in their own disciplines. By extracting data from a publicly available dataset, Dr. Fatih Aktas intended to use Gephi in the field of comparative and international education. This experience highlights the great potential of using network analysis in multidisciplinary research. The use of Gephi goes beyond the boundaries of social network analysis, network analysis can also become a tool for us as historians and humanities scholars — to go beyond the boundaries of our own expertise and engage with other disciplines.