Interpreting patterns through Social Network Analysis

Our map visualisation, launched in November 2023, allows users to explore the movements of over 3000 trainee African diplomats, using filters to provide a broad overview, or to ‘zoom in’ for a close reading. However, it does not provide a way to visualise or quantify network relationships, and for this we must turn to the Social Network Analysis techniques we piloted earlier. Treating the individual trainees as ‘nodes’ and their attendance on the various courses as ‘edges’, through the Gephi software, we constructed a non-spatial network visualisation that maps the social networks formed through the various courses and those that attended them.

As with the map visualisation, we are immediately able to interpret the main patterns in this data, spotting for instance its tendency towards mostly disconnected clusters representing the various cohorts of students. The centre of the network is dominated by IRIC students and their connections. It is also a more multilingual set of clusters than those disconnected ones on the periphery. Several of these patterns perceived visually can be tested by applying various statistical measures.

For example, ‘betweenness centrality’ is an indicator of a node’s centrality in a network, so running this test on our dataset helps to identify trainees who established some of the widest networks of contacts through their attendance of these courses. Six individuals stand out by an order of magnitude as having significantly higher ‘betweenness centrality’ (see table 1). They share the following traits: they are all francophone diplomats from West Africa, almost all of whom studied at IRIC in the mid-late 1970s. We can infer from this fact not only that the courses at IRIC were the most significant site for the networking of African diplomats, and in particular during the 1970s, but also that the greatest beneficiaries of this networking were not the Cameroonians hosting but their primarily francophone neighbours visiting, as they more often studied in one or more courses elsewhere as well. The highest ranking anglophone trainee, the late Paul Essel of Ghana, also attended IRIC in 1978 before taking a course at the IIAP in Paris in 1985.

Last Name	First Names	Country	Courses Attended	Betweenness Centrality (rank)	Combined Weighted Degree (rank)
DRAMANI	Dama Alfred	Togo	IRIC 1975, IIAP 1979	116467 (1)	285 (=30)
AGOSSOU	Albert	Benin	IIAP 1971, IRIC 1973, FSP 1985	100861 (2)	330 (13)
TIENDREBEOGO	Anatole	Burkina Faso	IRIC 1973, IRIC 1975	95108 (3)	397 (4)
BA	Bassirou	Mali	IHEOM 1965, UNITAR 1966, IIAP 1971	82629 (4)	240 (=53)
SANON	Pierre	Burkina Faso	IHEOM 1962, IRIC 1973	52851 (5)	176 (=103)
ESSEL	Paul	Ghana	IRIC 1978, IIAP 1985	17376 (13)	295 (24)

However, ‘betweenness centrality’ does not control for the added value of networking across different nationalities and language groups, which is an important consideration in relation to African Unity. To account for this attribute within the network analysis (following Borgatti and Halgin 2011), we assigned a numerical value (weights) to the relationships (edges) between trainees (nodes), based on the following criteria:

International Weight: same country (1), different country (2), new different country (4)

Interlingual Weight: same language (1), different language (2)

Combined Weight = International Weight x Interlingual Weight

Once these weights were assigned we used Gephi to calculate Weighted Degree (the sum of all edges’ weights attached to a given node) to highlight trainees that were particularly connected across national and linguistic borders through the training programmes. By this metric, francophone diplomats are comparatively less dominant and diplomats from small states such as Somalia, Zambia and Equatorial Guinea come to the fore. Attendees of IRIC and IIAP, which were more bilingual multinational courses, remain the highest scorers.

Last Name	First Names	Country	Courses Attended	Betweenness Centrality (rank)	Combined Weighted Degree (rank)
DIRIE	Ismail Mohamed	Somalia	IRIC 1971, IRIC 1972	6891 (=21)	564 (1)
MWABA	Michaël	Zambia	IRIC 1975	0 (=44)	412 (2)
IDOWU	Herekich Oludare	Nigeria	IIAP, 1970	0 (=44)	409 (3)
TIENDREBEOGO	Anatole	Burkina Faso	IRIC 1973, IRIC 1975	95108 (3)	397 (4)
MANGUE	Ela	Equatorial Guinea	IRIC 1975	0 (=44)	369 (5)

Visualising this high-volume, low-complexity data and using Social Network Analysis can only really be a first step. More nuance is needed, triangulating it with our other, more complex qualitative data. A combination of these measures helps us to select cases for ‘close reading’ through the (multi-)biographical approach we have taken throughout the project as a whole.