I have decided to start a new blog at https://philosophicalostrogoth.home.blog, as I am about to begin postgraduate studies. I will no longer update this site.
Friday, 19 July 2019
Tuesday, 18 June 2019
Dissertation Journal #5: Clusters and Conclusions
I have recently received my unconfirmed mark for my dissertation. I will therefore conclude my journal by discussing the final stages of my dissertation This will be done firstly be done by examining my thinking process for the final chapter and then by looking at potential directions for future research.
In my previous post, I discussed an anomaly surrounding the Roman Senate and how this helped me to think of an alternative model to the Gothic/Roman binary, which has often been used to understand social relations in Ostrogothic Italy. The final chapter of my dissertation aimed to take these ideas forward and apply them to the rest of the network. The first idea took forward from my previous chapter concerned the interaction between quantitative and qualitative techniques. In particular, how they are mutually beneficial to each other and the difficulty of using them in isolation. As mentioned earlier, my network was highly clustered and it was difficult to find a single 'rule' governing it. I therefore used Gephi's community detection features to isolate and identify small clusters, before examining the qualitative context from which they emerged from. The modularity algorithm used detected 96 communities in total.
I classified these communities under three main headings; 'administrative', 'diplomatic' and 'legal'. The intention of this was not to reduce all clusters into three types, but to provide a platform and structure for discussing the peculiarities of each individual instance. At this point, other ideas from the earlier in the dissertation came back to the forefront. Firstly, the overemphasis on ethnicity found in the current historiography. Secondly, the varying importance of alternative factors such as geographic location and particular titles. Finally, the role of 'practical reasons' or 'events' in causing connections. These ideas, with their emphasis on focusing on the microscopic level and the uniqueness of individual clusters, seemed to fit with the rest of the network.
Naturally, not every node or cluster followed these rules. A cluster surrounding two Goths, Duda and Tezutzat, protecting a Roman civilian called Petrus, reinforced rather than defied historiographic assumptions. A major surprise was that Praetorian Prefects, theoretically some of the most powerful individuals in Ostrogothic Italy, tended to have very few social connections. I suggested this was because they tended to move around quickly due to their wide remit over Italy. They struggled to build the more extensive networks of localised individuals, such as Urban Prefects. My emphasis in this chapter was on the need for a flexible understanding of social relations. One of my main issues with the Gothic/Roman binary is its simplicity and how historians (often, though not necessarily always) use it as a single answer for the social relations of an entire society.
A further issue addressed was the role of clustering in the network. Why was there no overarching rule to the network? Why were connections operating around separated and small communities? Following on from earlier ideas, I provided a number of explanations. Under the 'administrative' heading, I suggested clustering could be linked to Bjornlie's argument that Ostrogothic Italy had a decentralised and ad-hoc approach to governance. Individuals connecting around bureaucratic operations were chosen because they were in the right location at the right time and because they had the title necessary for them to carry out a task. There was not a central 'pool' of officials around one area, individuals were spread out across Italy. However, I was hesitant to use my data to support Bjornlie's further suggestion a more decentralised approach to governance was indicative of decline. Firstly, due to the lack of a comparative data sample prior to Ostrogothic rule in Italy. Secondly, because a decentralised approach may have been a conscious and proactive decision to make the administration more suitable for the early sixth-century, rather than a simple fall from earlier Roman standards. In the context of 'legal' clusters, this explanation for clustering was again relevant. Lafferty has argued legal cases were dealt with by a provincial governor, irrespective of an individuals' ethnicity. Therefore, clusters surrounding law were also linked to a 'ad-hoc' decentralised approach. Finally, I also made some indications that the temporariness of connections may account for some of the clustering. If individuals did not 'socialise' or 'connect' outside of events or government orders, they would not have had any opportunities to develop more complete networks.
In my previous post, I discussed an anomaly surrounding the Roman Senate and how this helped me to think of an alternative model to the Gothic/Roman binary, which has often been used to understand social relations in Ostrogothic Italy. The final chapter of my dissertation aimed to take these ideas forward and apply them to the rest of the network. The first idea took forward from my previous chapter concerned the interaction between quantitative and qualitative techniques. In particular, how they are mutually beneficial to each other and the difficulty of using them in isolation. As mentioned earlier, my network was highly clustered and it was difficult to find a single 'rule' governing it. I therefore used Gephi's community detection features to isolate and identify small clusters, before examining the qualitative context from which they emerged from. The modularity algorithm used detected 96 communities in total.
I classified these communities under three main headings; 'administrative', 'diplomatic' and 'legal'. The intention of this was not to reduce all clusters into three types, but to provide a platform and structure for discussing the peculiarities of each individual instance. At this point, other ideas from the earlier in the dissertation came back to the forefront. Firstly, the overemphasis on ethnicity found in the current historiography. Secondly, the varying importance of alternative factors such as geographic location and particular titles. Finally, the role of 'practical reasons' or 'events' in causing connections. These ideas, with their emphasis on focusing on the microscopic level and the uniqueness of individual clusters, seemed to fit with the rest of the network.
Graph showing size of clusters in the network.
Naturally, not every node or cluster followed these rules. A cluster surrounding two Goths, Duda and Tezutzat, protecting a Roman civilian called Petrus, reinforced rather than defied historiographic assumptions. A major surprise was that Praetorian Prefects, theoretically some of the most powerful individuals in Ostrogothic Italy, tended to have very few social connections. I suggested this was because they tended to move around quickly due to their wide remit over Italy. They struggled to build the more extensive networks of localised individuals, such as Urban Prefects. My emphasis in this chapter was on the need for a flexible understanding of social relations. One of my main issues with the Gothic/Roman binary is its simplicity and how historians (often, though not necessarily always) use it as a single answer for the social relations of an entire society.
A further issue addressed was the role of clustering in the network. Why was there no overarching rule to the network? Why were connections operating around separated and small communities? Following on from earlier ideas, I provided a number of explanations. Under the 'administrative' heading, I suggested clustering could be linked to Bjornlie's argument that Ostrogothic Italy had a decentralised and ad-hoc approach to governance. Individuals connecting around bureaucratic operations were chosen because they were in the right location at the right time and because they had the title necessary for them to carry out a task. There was not a central 'pool' of officials around one area, individuals were spread out across Italy. However, I was hesitant to use my data to support Bjornlie's further suggestion a more decentralised approach to governance was indicative of decline. Firstly, due to the lack of a comparative data sample prior to Ostrogothic rule in Italy. Secondly, because a decentralised approach may have been a conscious and proactive decision to make the administration more suitable for the early sixth-century, rather than a simple fall from earlier Roman standards. In the context of 'legal' clusters, this explanation for clustering was again relevant. Lafferty has argued legal cases were dealt with by a provincial governor, irrespective of an individuals' ethnicity. Therefore, clusters surrounding law were also linked to a 'ad-hoc' decentralised approach. Finally, I also made some indications that the temporariness of connections may account for some of the clustering. If individuals did not 'socialise' or 'connect' outside of events or government orders, they would not have had any opportunities to develop more complete networks.
Having outlined some of my conclusions during the latter stages of my dissertation, I will now discuss some of the potential directions for future research. The first would be to simply use more varied and extensive evidence to further test my findings. My dissertation covered the first books of Cassiodorus' Variae, which are all from the reign of Theoderic the Great. Using the latter books and other sources, like Procopius' History of the Wars, would allow an analysis that focuses more heavily on any potential changes over time. Dynamic Network Analysis (DNA) rather than standard SNA would be more suitable for this task, as it represents change better than the static visualisations and metrics used in my dissertation. I also think consulting other types of sources would be useful. Even Boethius' Consolation of Philosophy contains incidental references to social connections, considering other genres like this could be useful. They could offer further insight due to their differences to the government-oriented Variae, in particular regarding connections outside of 'practical' concerns or events.
Aside from confirming or refuting my findings, I believe my ideas could be took into another direction. Network analysis could be used to examine a number of interdisciplinary and philosophical issues. Many of the findings in my dissertation indicate it would be wrong to view Ostrogothic social relations as 'simple', instead they seem to match with the concept of 'complexity' found in the natural and social sciences. I believe there is room for interdisciplinary dialogue here, particularly regarding the ontological nature of relations. By 'borrowing' Network Analysis for a historical study, I have portrayed relations as polyadic. This would seem to be at odds with the binarism prevalent in the current historiography on Goths/Romans. Several interesting questions could be raised here. How far is my representation of relations an accurate reflection of reality or alternatively a result of using a technique originally developed in other disciplines? Likewise, if Late Antique philosophers conceived of relations generally in a non-polyadic fashion, would there be a problematic contradiction with my findings? Finally, there is also a need to consider the presumptions historians are carrying when analysing relations. By suggesting Goths/Romans were separate, yet harmoniously working together, are we not be presuming the subjects of a relation have greater ontological priority than the relation itself? I believe these theoretical issues could be addressed by using Network Analysis as a 'gateway' for relevant discussion.
Having concluded my dissertation, I believe there is room to develop some of my ideas. By keeping a journal, I have hoped to keep a record of the processes involved in writing my dissertation. Firstly, for my own sake, in case I return to these ideas in the future. Secondly, because I genuinely believe some of the ideas contained in my dissertation could be quite important if developed further. Overall, keeping a journal has been an enjoyable process and I hope my entries have given some insight into my thinking.
Aside from confirming or refuting my findings, I believe my ideas could be took into another direction. Network analysis could be used to examine a number of interdisciplinary and philosophical issues. Many of the findings in my dissertation indicate it would be wrong to view Ostrogothic social relations as 'simple', instead they seem to match with the concept of 'complexity' found in the natural and social sciences. I believe there is room for interdisciplinary dialogue here, particularly regarding the ontological nature of relations. By 'borrowing' Network Analysis for a historical study, I have portrayed relations as polyadic. This would seem to be at odds with the binarism prevalent in the current historiography on Goths/Romans. Several interesting questions could be raised here. How far is my representation of relations an accurate reflection of reality or alternatively a result of using a technique originally developed in other disciplines? Likewise, if Late Antique philosophers conceived of relations generally in a non-polyadic fashion, would there be a problematic contradiction with my findings? Finally, there is also a need to consider the presumptions historians are carrying when analysing relations. By suggesting Goths/Romans were separate, yet harmoniously working together, are we not be presuming the subjects of a relation have greater ontological priority than the relation itself? I believe these theoretical issues could be addressed by using Network Analysis as a 'gateway' for relevant discussion.
Having concluded my dissertation, I believe there is room to develop some of my ideas. By keeping a journal, I have hoped to keep a record of the processes involved in writing my dissertation. Firstly, for my own sake, in case I return to these ideas in the future. Secondly, because I genuinely believe some of the ideas contained in my dissertation could be quite important if developed further. Overall, keeping a journal has been an enjoyable process and I hope my entries have given some insight into my thinking.
Primary Sources:
Boethius, De Consolatione Philosophiae in The Consolation of Philosophy translated
by Victor Watts. London: Penguin, 1969.
Cassiodorus,
Variae in The Letters of
Cassiodorus: Being A Condensed Translation Of The Variae Epistolae Of
Magnus Aurelius Cassiodorus Senator translated by Thomas Hodgkin. London:
Henry Frowde, 1886.
Procopius,
History of the Wars in Procopius translated by William H. Dewey in 7 volumes. Cambridge, MA: Harvard
University Press, 1914-1940.
Secondary Sources:
Bjornlie,
Shane. "Governmental Administration." In A Companion to Ostrogothic Italy, edited by
Jonathan J. Arnold, Shane Bjornlie and Kristina Sessa, 47-72. Leiden: Brill,
2016.
Carley,
Kathleen M. "Dynamic Network Analysis." In Dynamic Social Network Modeling and Analysis: Workshop Summary
and Papers, edited by Ronald Breiger, Kathleen M. Carley and Philippa
Pattison, 133-45. Washington: National Academies Press, 2003.
Lafferty,
Sean D.W. Law and Society in the Age of
Theoderic: A Study of the Edictum Theoderici. Cambridge: Cambridge University Press, 2013.
-
Thursday, 28 March 2019
Dissertation Journal #4: Explaining an Anomaly and Methodological Issues
Since my last post, I have completed the first draft of my dissertation. In this entry, I hope to raise some of the methodological issues that I encountered during this process. This will be done by focusing on an anomaly in my data that centered around the Roman Senate and the problems I faced while trying to explain it through Social Network Analysis (SNA).
I will firstly highlight the characteristics that separated the anomaly from the norms of the network. Firstly, it was a relatively dense and highly-connected area of the network. The findings of my first chapter suggested that Ostrogothic Italy on the whole tended to be a relatively divided society. However, the anomalous group of 39 nodes challenged this. The betweenness centrality for this cluster was 8.946078412, whereas for the entire network it was only 1.164930545 (Theoderic the Great was excluded from these calculations as he connects to everyone). A higher score on this metric indicates the extent to which a node mediates relations, so it became clear this group of 39 nodes had unusually high levels of interconnectivity. The second defining characteristic of the anomaly was that it had a higher percentage of nodes with the label 'Roman'. Non-Romans were only 10.24% of the cluster, in contrast to forming 51.9% of the entire network.
The first questions needing to be addressed were simple. Was this anomaly contradicting the points I developed earlier in the dissertation about ethnicity not being important in Ostrogothic social relations? Furthermore, could the anomaly instead be characterised as an 'ethnic' enclave for Romans? The answers were not necessarily simple. While being numerically inferior, 'Gothic' and 'Unknown' nodes still tended to score equally if not higher than 'Roman' nodes on most metrics. In other words, this anomaly was an intensification of the patterns established in my first chapter. There were more Romans, but this did not necessarily translate to them being more important or separate from Goths in terms of connections.
By examining other factors that may have influenced the creation of the anomaly, I hoped to more adequately address this group of nodes. I firstly assessed if holding a certain 'title' was important. Firstly, I noticed the anomaly had a high concentration of individuals who were Viri Illustres. By Late Antiquity, this was the only title that allowed admission to the Roman Senate. It is worth noting here that during this period an individual could very well hold senatorial status through a different title, but only this one could allow participation in the Senate as a political institution. However, aside from this, there were also some odd titles in the data, such as Arch-Physician and Pantomimist. In this way, it was clear that participation in the Senate was important, but at the same time more answers were needed. An analysis of individuals' zones of activity yielded similar results. Rome was particularly prominent as a location for individuals in this group, which again reinforced my conviction that the Senate was playing an important role. With this mind, I suggested the higher rates of interconnectivity in the anomaly could potentially be linked to the the independence the Senate still held as an institution, with it being less dependent on the patronage of Theoderic. Meanwhile, the changes in the ratio of Goths to Romans were a result of the Senate's relatively small size in comparison to the rest of the Ostrogothic bureaucracy. It was harder for everyone, not just Goths, to be a part of the Senate. Despite all of this, there were still too many nodes within the anomaly that could not be entirely explained by the Roman Senate and so I was required to consider additional factors.
A screenshot of part of the anomaly, with Romans in red and the other nodes representing nodes with non-Roman labels.
It was at this point, I began to consider a number of methodological issues. The first was whether my application of SNA to the letters of Cassiodorus was giving undue importance to this anomaly. This came to mind because the group contained a particularly prominent clique originating from only 2 of the 229 letters studied. A clique is a set of nodes where every possible connection is made. The connections in this clique occurred twice, which was highly unusual, as individuals in the network tended to connect only once and never again. However, there were only 9 nodes in this clique and 30 other ones with the aforementioned anomalous traits still remained. Further answers were needed.
Having exhausted what the data was telling me about the anomaly, I decided to see if I was missing something by using quantitative rather than qualitative techniques. In particular, I wanted to examine the contexts in which individual connections were taking place. By doing this, I noticed that individuals were primarily meeting as a response to 'events' or for practical reasons, such as an administrative order (e.g collecting taxes, etc). This allowed me to develop a model for understanding the anomaly which relied on both SNA and readings of my source. Factors, such as 'titles' or 'zones of activity' were indeed important in the creation of the anomaly. However, social relations are not necessarily predictable. For example, we cannot presume the pantomimist Helladius would have not had everyday access to the powerful Urban Prefect of Rome. Yet, they connected anyway due to disturbances at the circus. It was on this basis, I suggested that while certain factors can influence the likelihood of a connection (in this case, they both shared Rome), they do not in themselves create it- there needs to be a reason for a relation. In this way, the quantitative data was useful for identifying that being part of a Roman Senate was an important factor for the anomaly, whereas looking at the sources allowed a more fuller explanation for any peculiarities. The Roman Senate was not rigid, it encountered people outside of its own ranks due to daily practical activities, and to expect a 'pure' undifferentiated set of nodes within the anomaly would have missed the point. It would have portrayed social relations as hierarchical, predictable and overly simplistic.
Following this necessary interaction of quantitative and qualitative techniques, I thought it was necessary to raise some of the methodological implications. For example, are statistics better at explaining some problems better than others? Should the historian use SNA and traditional document-reading alongside each other? Based on my findings, I suggested it would be wrong to take the view we need to choose one or the other. In fact, both methodologies can be mutually beneficial to each other. SNA can be helpful for complicating structures that may have been taken for granted in the historiography (in this instance, the role of ethnicity), while also allowing us to assess the importance of other factors in social relations. Whereas, qualitative techniques allow a more microscopic look at instances which defy our expectations. I took many of the ideas developed when analysing this anomaly into my final chapter, which looked at the variety of smaller clusters that made up the rest of the network. Hopefully, this post offers insight into my thought processes on the methodological issues I encountered while writing my dissertation.
Having exhausted what the data was telling me about the anomaly, I decided to see if I was missing something by using quantitative rather than qualitative techniques. In particular, I wanted to examine the contexts in which individual connections were taking place. By doing this, I noticed that individuals were primarily meeting as a response to 'events' or for practical reasons, such as an administrative order (e.g collecting taxes, etc). This allowed me to develop a model for understanding the anomaly which relied on both SNA and readings of my source. Factors, such as 'titles' or 'zones of activity' were indeed important in the creation of the anomaly. However, social relations are not necessarily predictable. For example, we cannot presume the pantomimist Helladius would have not had everyday access to the powerful Urban Prefect of Rome. Yet, they connected anyway due to disturbances at the circus. It was on this basis, I suggested that while certain factors can influence the likelihood of a connection (in this case, they both shared Rome), they do not in themselves create it- there needs to be a reason for a relation. In this way, the quantitative data was useful for identifying that being part of a Roman Senate was an important factor for the anomaly, whereas looking at the sources allowed a more fuller explanation for any peculiarities. The Roman Senate was not rigid, it encountered people outside of its own ranks due to daily practical activities, and to expect a 'pure' undifferentiated set of nodes within the anomaly would have missed the point. It would have portrayed social relations as hierarchical, predictable and overly simplistic.
Following this necessary interaction of quantitative and qualitative techniques, I thought it was necessary to raise some of the methodological implications. For example, are statistics better at explaining some problems better than others? Should the historian use SNA and traditional document-reading alongside each other? Based on my findings, I suggested it would be wrong to take the view we need to choose one or the other. In fact, both methodologies can be mutually beneficial to each other. SNA can be helpful for complicating structures that may have been taken for granted in the historiography (in this instance, the role of ethnicity), while also allowing us to assess the importance of other factors in social relations. Whereas, qualitative techniques allow a more microscopic look at instances which defy our expectations. I took many of the ideas developed when analysing this anomaly into my final chapter, which looked at the variety of smaller clusters that made up the rest of the network. Hopefully, this post offers insight into my thought processes on the methodological issues I encountered while writing my dissertation.
Secondary Sources:
Heather, Peter. "Senators and Senates." In The Cambridge Ancient History: Volume 13: The Late Empire, AD 337–425, edited by Averil Cameron and Peter Garnsey, 184-210. Cambridge: Cambridge University Press, 1997.
Radtki, Christine. "The Senate at Rome in Ostrogothic Italy." In A Companion to Ostrogothic Italy, edited by Jonathan J. Arnold, Shane Bjornlie and Kristina Sessa, 121-46. Leiden: Brill, 2016.
Scott, John. Social Network Analysis: A Handbook. 2017 ed. Thousand Oaks, California: SAGE Publications, 1991.
Wednesday, 23 January 2019
Dissertation Journal #3: Applying Metrics and Analysing the Network
In my last post, I discussed the processes involved in constructing a social network. This entry will build on this by looking at how I analysed the network. I will firstly introduce some of the methods used to do this, before discussing what they reveal about social relations in Ostrogothic Italy. Most of what will be discussed here ended up forming the basis of my the first chapter of my dissertation and so this entry will give some insight into how I came to my conclusions.
The software used for analysing the network was Gephi, which aids Social Network Analysis (SNA) in two ways. Firstly, by visualising the network, allowing the researcher to identify patterns. An image of part of my network, as visualised in Gephi, can be seen below. To aid my investigations for the first chapter, I colour-coded the different nodes based on their 'ethnic' label. Red nodes were Roman, whereas purple ones were Gothic. Likewise, blue nodes represented individuals that are classified as 'Unknown' and green ones represent individuals classified under 'Other', such as Visigoths or Franks. In this image, lines between people mean they are connected at least once. A visual representation such as this was useful, as it allowed me to analyse the network on a smaller scale. For example, by looking at a representation, I was able to identify a clique based around the Roman Senate going against the norm of the network established by metrical means.
Part of the network, colour-coded for my analysis.
I will now move on to the aforementioned ability to use metrics in Gephi. Social Network Analysis often involves using a range of formula which help the researcher to identify patterns. These can be calculated for the network as a whole. For example, global density helps to identify how complete or connected the entirety of the network is. This is calculated by expressing the number of lines in the network as a proportion of the maximum number of lines. It was therefore useful for identifying the divided nature of my network. The average clustering coefficient is another metric calculated globally and is used to understand the tendency of individuals to form into clusters within the network.
There are also a number of metrics that assign scores to individuals, rather than on a global-level. Closeness centrality calculates the shortest route through which a node tends to be connected to other nodes in the network, whether this is directly or through an intermediary. Eigenvector centrality is similar to this, except it is based on how far a node is connected to the most central nodes. Metrics such as these can be useful for finding out who the most important people in the network are. However, they are also useful for finding out averages for different node types. For example, it was possible to work out that Goths and Romans were generally equal in importance in terms of their place in the network. As I used a range of technical vocabulary, like the ones described here, throughout the first chapter, I decided to restrict definitions to the appendices of my dissertation. This is to prevent detraction from my argument in the main body of the text.
I will now discuss some of my findings from applying these forms of analyses for the first chapter of my dissertation. The first and most significant finding was that connections between individuals did not tend to be made in the context of ethnicity. Most modern scholarship accepts that ethnicity does not have an unchanging biological basis. However, historians have still tended to emphasise the importance of ethnicity, even it its ideological or constructed form, when understanding Ostrogothic society. My research has questioned this assumption in multiple ways. Firstly, as mentioned, average centrality scores for individuals in the network tended to be the same for Goths and Romans. They were unable to be distinguished in this way. Secondly, based on a visual analysis of my network I noticed that while nodes did tend to coalesce into groups, there was no basis to use ethnicity as an explanation for this. Nodes of different ethnic labels tended to connect with each other more often than not. The final statistic that supported this point were the percentages at which Goths and Romans connected with other 'ethnic' labels. Based on a 20% sample of all Goths in the network, I found out that Goths tended to connect more with Romans rather than their fellow Goths.
The second finding of my research was that the military/civilian divide that has characterised discussion around Ostrogothic Italy is not supported by a statistical analysis. Goths have often been seen as 'soldiers' and Romans as 'civilians'. Nearly every single title within my study was either civilian or straggled the apparent divide. For example, many titles such as Dukes, had civilian and military duties. However, some titles did seem to be associated with particular 'ethnic' labels. For example the 17 Sajones in the data were all Gothic. Sajones were experienced soldiers, who usually held judicial roles. However, this was not a suitable foundation to argue for a military/civilian divide, the holders of this title still tended to associate with Romans more than Goth, in spite of being Gothic themselves.
This entry has established the different methodologies I used to carry out an analysis of social relations in Ostrogothic Italy and pointed towards how I came to the conclusions established in the first chapter of my dissertation. After dismissing the traditional ethnic interpretations, the rest of the analyses of my network will focus on establishing a new model for understanding society in early sixth-century Italy.
Secondary Sources:
The software used for analysing the network was Gephi, which aids Social Network Analysis (SNA) in two ways. Firstly, by visualising the network, allowing the researcher to identify patterns. An image of part of my network, as visualised in Gephi, can be seen below. To aid my investigations for the first chapter, I colour-coded the different nodes based on their 'ethnic' label. Red nodes were Roman, whereas purple ones were Gothic. Likewise, blue nodes represented individuals that are classified as 'Unknown' and green ones represent individuals classified under 'Other', such as Visigoths or Franks. In this image, lines between people mean they are connected at least once. A visual representation such as this was useful, as it allowed me to analyse the network on a smaller scale. For example, by looking at a representation, I was able to identify a clique based around the Roman Senate going against the norm of the network established by metrical means.
I will now move on to the aforementioned ability to use metrics in Gephi. Social Network Analysis often involves using a range of formula which help the researcher to identify patterns. These can be calculated for the network as a whole. For example, global density helps to identify how complete or connected the entirety of the network is. This is calculated by expressing the number of lines in the network as a proportion of the maximum number of lines. It was therefore useful for identifying the divided nature of my network. The average clustering coefficient is another metric calculated globally and is used to understand the tendency of individuals to form into clusters within the network.
There are also a number of metrics that assign scores to individuals, rather than on a global-level. Closeness centrality calculates the shortest route through which a node tends to be connected to other nodes in the network, whether this is directly or through an intermediary. Eigenvector centrality is similar to this, except it is based on how far a node is connected to the most central nodes. Metrics such as these can be useful for finding out who the most important people in the network are. However, they are also useful for finding out averages for different node types. For example, it was possible to work out that Goths and Romans were generally equal in importance in terms of their place in the network. As I used a range of technical vocabulary, like the ones described here, throughout the first chapter, I decided to restrict definitions to the appendices of my dissertation. This is to prevent detraction from my argument in the main body of the text.
I will now discuss some of my findings from applying these forms of analyses for the first chapter of my dissertation. The first and most significant finding was that connections between individuals did not tend to be made in the context of ethnicity. Most modern scholarship accepts that ethnicity does not have an unchanging biological basis. However, historians have still tended to emphasise the importance of ethnicity, even it its ideological or constructed form, when understanding Ostrogothic society. My research has questioned this assumption in multiple ways. Firstly, as mentioned, average centrality scores for individuals in the network tended to be the same for Goths and Romans. They were unable to be distinguished in this way. Secondly, based on a visual analysis of my network I noticed that while nodes did tend to coalesce into groups, there was no basis to use ethnicity as an explanation for this. Nodes of different ethnic labels tended to connect with each other more often than not. The final statistic that supported this point were the percentages at which Goths and Romans connected with other 'ethnic' labels. Based on a 20% sample of all Goths in the network, I found out that Goths tended to connect more with Romans rather than their fellow Goths.
The second finding of my research was that the military/civilian divide that has characterised discussion around Ostrogothic Italy is not supported by a statistical analysis. Goths have often been seen as 'soldiers' and Romans as 'civilians'. Nearly every single title within my study was either civilian or straggled the apparent divide. For example, many titles such as Dukes, had civilian and military duties. However, some titles did seem to be associated with particular 'ethnic' labels. For example the 17 Sajones in the data were all Gothic. Sajones were experienced soldiers, who usually held judicial roles. However, this was not a suitable foundation to argue for a military/civilian divide, the holders of this title still tended to associate with Romans more than Goth, in spite of being Gothic themselves.
This entry has established the different methodologies I used to carry out an analysis of social relations in Ostrogothic Italy and pointed towards how I came to the conclusions established in the first chapter of my dissertation. After dismissing the traditional ethnic interpretations, the rest of the analyses of my network will focus on establishing a new model for understanding society in early sixth-century Italy.
Secondary Sources:
Amory, Patrick. People and Identity in Ostrogothic Italy, 489-554. Cambridge; New York: Cambridge University Press, 1997.
Heather, Peter. "Merely an Ideology? Gothic Identity in Ostrogothic Italy." In The Ostrogoths from the Migration Period to the Sixth Century: An Ethnographic Perspective, edited by Sam Barnish and Federico Marazzi, 31-80. Woodbridge, Suffolk: Boydell Press, 2007.
Knoke, David and Song Yang. Social Network Analysis. Thousand Oaks: SAGE Publications, 2000.
Scott, John. Social Network Analysis: A Handbook. 2017 ed. Thousand Oaks, California: SAGE Publications, 1991.
Subscribe to:
Posts (Atom)