Commetrix - IRIS co-authorship network analysis

CASE: IRIS (1978-2006) Historical Reflection through Visual Analysis
ONLINE SUPPLEMENT of the paper by Molka-Danielsen, J., Trier, M., Shlyk, V., Bobrik, A., Nurminen, M.I. (2007): IRIS (1978-2006) Historical Reflection through Visual Analysis. Proceedings of IRIS30, Tampere, August 2007, Finland. Paper as PDF Presentation Slides as PDF . . INTRODUCTION. IRIS (Information Systems Research in Scandinavia) Conference is an annual meeting for researchers in Scandinavia . Through 29 annual conference meetings a social network of co-authorships and general research collaboration has emerged. This article contributes a historical reflection and analysis of the social network of researchers associated with the IRIS in the period from 1978 to 2006. Using Commetrix, we approach the following research questions: aggregate general statistics of the co-authorship networks of IRIS visualize a map of co-authorship network after 30 years of IRIS and observe its development over time identify and visualize the most prominent actors and their ego-networks capture the nature and characteristic of the research topics over time and visualize the network of keywords found in the conference paper titles and visualizing their representation in the network. This supplement contains large versions of the paper's figures plus additional visual material like animations as movies. A mouse click on each thumbnail opens the related ressource. Right mouse button can be used to save the ressource locally. . . *FIGURE 2 . The complete author and co-author relationship network. node: author \| link: co-authorship of adjacent authors \| node color: groups \| node size: contributed paper of author \| distance: strength of association between two authors \| node label: author-ID . Observation. This figure shows the over picture of the IRIS co-author network from 1978 to 2006. Different colors represent different groups. 1252 participants of the IRIS conferences that simply attended or submitted at least one paper. Of the 1252 participants there are 1219 authors of papers (this number of authors is shown in the picture, i.e. also individual authorsand as well as 976 of these are co-authors. Participating as a co-author on a paper creates links in the network. There is a large group (biggest component) and many smaller groups. The network graph shows one large component of 541 linked authors (44%) and several smaller clusters with the second largest cluster only containing 23 authors (2%). Over time, the 44% authors of the large component have produced a very interrelated and large researcher network. More peripheral co-authorships, which do not belong to this central cluster appear as small network components, meaning that they remained much more isolated and less complex. About 10 actors appear very active (large nodes) in this network. Looking at their embeddedness it can be seen, that those active authors are also the more central persons in the IRIS network. . . . FIGURE 3. Subset of 256 actors with a minimum of three contributed papers. node: author \| link: co-authorship of adjacent authors \| node color: activity in terms of contributed papers \| node size: number of co-authorship contacts of author \| distance: strength of association between any two authors \| node label: author-ID . Observation. This threshold improves transparency by decreasing the number of shown nodes. The sample now has 440 relationships, a density of 1.35% and an average degree = 4,182 links per author. At the core sof the IRIS co-authorship network a central axis emerges with nodes 12, 58, 1517, 160, and 187 at the center. A second pattern appears as long paths reaching out towards the periphery of the network and as very large (intertwined) rings with more than 13 steps. The authors positioned in these structures form large chains of peer relationships. . . . FIGURE 4. The different states of IRIS network in 4 periods. node: author \| link: co-authorship of adjacent authors \| node size: activity in terms of contributed papers \| node color: number of co-authorship contacts of author \| distance: strength of association between any two authors \| node label: author-ID . Observation. Using a time window, the relationships outside the period are ignored to better capture the actual network that has been active at each period. There is one initial cluster in period 1, which slowly dissolves. However, a second cluster organically emerges in the network and builds througout periods 2 and 3 to become the dominant center of the co-authorship network. Only one node of the initial cluster in period 1 gets embedded in this new center and develops to an integral part of it. In period four the center moves to a new most central actor which had its first co-authorship in 1997. . . . FIGURE 4a - MOVIE. The evolving IRIS network (change only, with time window of 25% of overall period - links can decay and disappear) (40MB Quicktime .mov). node: author \| link: co-authorship of adjacent authors \| node size: activity in terms of contributed papers \| node color: number of co-authorship contacts of author \| distance: strength of association between any two authors \| node label: author-ID . Observation. This video demonstrates the co-authorship network of IRIS and its development over time. Different dominating clusters can be observed and the changes of central areas discussed in the previos figure. The time window is set to 25% of overall period. This eliminates edges (and according nodes) from the graph that are too old and highlights the change of the network. . . . FIGURE5. Combined ego-network among the top 5 central authors. node: author \| link: co-authorship of adjacent authors \| node color: top 5 most central nodes (egos) \| node size: number of co-authorship contacts of author in visible combined ego-network \| distance: strength of association between any two authors \| node label: author-ID . Observation. The lighter color nodes in Figure 5 represent the top five central actors selected as ego. Interestingly, four further actors of the top6-top10 actors in Table 4 are also found in direct relationship to these top5. The peers of the top 5 ego nodes did not develop short cut links but are only linked indirectly via the ego nodes. This means, there is no strong effect of triadic closure. . . . FIGURE6. Ego-network among the top 5 central authors first vs. last years. node: author \| link: co-authorship of adjacent authors \| node color: top 5 most central nodes (egos) \| node size: number of co-authorship contacts of author in visible combined ego-network \| distance: strength of association between any two authors \| node label: author-ID . Observation. This figure separates the first half until 1991 and the second half of the sampling period. the development of the top 5 central nodes corresponds to the general development of the IRIS network and its shifting epicenters throughout the analyzed four time periods. Node 12 was very central during the first period. Later node 58 moved to the central position in period two. In this period, node 160 was embedded in a very small peripheral component, but it moved towards a very central position in period three (starting in 1993). The same applies for node 213, except that this author started its centralization efforts later in 1998. Node 343 finally is the most connected node in the fourth period although the first co-authorship link was established only in 1997. For the observed IRIS network, this implies that the active centers of the overall structure are formed by very active and prominent authors during their lifecycle of activity. Although three out of the top five have even collaborated directly, their respective groups of co-authors remain very separated. . . . FIGURE6a. Combined ego-network among the top 5 active authors. node: author \| link: co-authorship of adjacent authors \| node color: top 5 most central nodes (egos) \| node size: number of co-authorship contacts of author in visible combined ego-network \| distance: strength of association between any two authors \| node label: author-ID . Observation. Altogether, this subnetwork includes 84 authors in 198 relationships. The lighter nodes represents those top 5 central persons. Interestingly, only two of the top6-top10 active actors are found in direct relationship to the top5. The rest are positioned in different sectors of the network meaning that several independent collaborations arise. . . . FIGURE6b. Ego-network among the top 5 active authors first vs. last years. node: author \| link: co-authorship of adjacent authors \| node color: top 5 most central nodes (egos) \| node size: number of co-authorship contacts of author in visible combined ego-network \| distance: strength of association between any two authors \| node label: author-ID . Observation. Some connections between ego-networks were created. However, at other points the top5 actors disconnected. In both periods we found three (different) clusters and thus the connected network has never really existed but emerges because of the accumulative nature of SNA graphs (i.e. there is no decay). We can conclude that the most central nodes where establishing their prominent position largely independently of each other. . . . FIGURE7. The diffusion of the terms information and knowledge over time. node: author \| link: co-authorship of adjacent authors \| node size: number of co-authorship contacts of author in visible combined ego-network \| distance: strength of association between any two authors \| node label: author-ID . Observation. The co-authorship network of authors, who employed the terms information and knowledge in their papers at three different points in time. This is indicating the spread of these terms over time. Note the different size of the network and its dispersion into many smaller subnetworks. . . . FIGURE 7a - MOVIE. The diffusion of the terms information over time (5MB Quicktime .mov). node: author \| link: co-authorship of adjacent authors \| node size: number of co-authorship contacts of author in visible combined ego-network \| distance: strength of association between any two authors \| node label: author-ID . Observation. The same co-authorship network of authors, who employed the term information in their papers, indicating the spread of these terms over time. This time, the animation can be observed. . . . FIGURE 7b - MOVIE. The diffusion of the terms knowledge over time (3MB Quicktime .mov). node: author \| link: co-authorship of adjacent authors \| node size: number of co-authorship contacts of author in visible combined ego-network \| distance: strength of association between any two authors \| node label: author-ID . Observation. The same co-authorship network of authors, who employed the term knowledge in their papers, indicating the spread of these terms over time. This time, the animation can be observed. . . . FIGURE 7c. Line Charts of keyword occurrence (trends) across time. color: each of the top 20 keywords. . Observation.** The absolute occurrence shows the constant occurrences of information, system, and development. Some terms like process, mobile, management, and knowledge increased in absolute occurrence. The percentual share of occurrence of those keywords (i.e. percent of papers with selected keyword versus overall papers) decreases noticeably over time. This highlights reduced coverage of the selected top twenty and indicates increased pluralism and heterogeneity: The few initial topics differentiate. . . . More related research can be found at: http://www.ikmresearch.de . *Numbering of Images corresponds to the Figure in the published paper. Additional material is marked with a subsequent letter, e.g. 3a)

Matthias Trier, © IKM/sysedv Technical University Berlin 2012

CASE: IRIS (1978-2006) Historical Reflection through Visual Analysis

ONLINE SUPPLEMENT
of the paper by Molka-Danielsen, J., Trier, M., Shlyk, V., Bobrik, A., Nurminen, M.I. (2007): IRIS (1978-2006) Historical Reflection through Visual Analysis. Proceedings of IRIS30, Tampere, August 2007, Finland.
Paper as PDF Presentation Slides as PDF

INTRODUCTION. IRIS (Information Systems Research in Scandinavia) Conference is an annual meeting for researchers in Scandinavia . Through
29 annual conference meetings a social network of co-authorships and general research collaboration has emerged. This article contributes a historical reflection and analysis of the social network of researchers associated with the IRIS in the period from 1978 to 2006. Using Commetrix, we approach the following research questions:

aggregate general statistics of the co-authorship networks of IRIS
visualize a map of co-authorship network after 30 years of IRIS and observe its development over time
identify and visualize the most prominent actors and their ego-networks
capture the nature and characteristic of the research topics over time and visualize the network of keywords found in the conference paper titles and visualizing their representation in the network.

This supplement contains large versions of the paper's figures plus additional visual material like animations as movies. A mouse click on each thumbnail opens the related ressource. Right mouse button can be used to save the ressource locally.

FIGURE 2* . The complete author and co-author relationship network.

Observation. This figure shows the over picture of the IRIS co-author network from 1978 to 2006. Different colors represent different groups. 1252 participants of the IRIS conferences that simply attended or submitted at least one paper. Of the 1252 participants there are 1219 authors of papers (this number of authors is shown in the picture, i.e. also individual authorsand as well as 976 of these are co-authors. Participating as a co-author on a paper creates links in the network. There is a large group (biggest component) and many smaller groups. The network graph shows one large component of 541 linked authors (44%) and several smaller clusters with the second largest cluster only containing 23 authors (2%). Over time, the 44% authors of the large component have produced a very interrelated and large researcher network. More peripheral co-authorships, which do not belong to this central cluster appear as small network components, meaning that they remained much more isolated and less complex. About 10 actors appear very active (large nodes) in this network. Looking at their embeddedness it can be seen, that those active authors are also the more central persons in the IRIS network.

FIGURE 3. Subset of 256 actors with a minimum of three contributed papers.

Observation. This threshold improves transparency by decreasing the number of shown nodes. The sample now has 440 relationships, a density of 1.35% and an average degree = 4,182 links per author. At the core sof the IRIS co-authorship network a central axis emerges with nodes 12, 58, 1517, 160, and 187 at the center. A second pattern appears as long paths reaching out towards the periphery of the network and as very large (intertwined) rings with more than 13 steps. The authors positioned in these structures form large chains of peer relationships.

FIGURE 4. The different states of IRIS network in 4 periods.

Observation. Using a time window, the relationships outside the period are ignored to better capture the actual network that has been active at each period. There is one initial cluster in period 1, which slowly dissolves. However, a second cluster organically emerges in the network and builds througout periods 2 and 3 to become the dominant center of the co-authorship network. Only one node of the initial cluster in period 1 gets embedded in this new center and develops to an integral part of it. In period four the center moves to a new most central actor which had its first co-authorship in 1997.

FIGURE 4a - MOVIE. The evolving IRIS network (change only, with time window of 25% of overall period - links can decay and disappear)
(40MB Quicktime .mov).

Observation. This video demonstrates the co-authorship network of IRIS and its development over time. Different dominating clusters can be observed and the changes of central areas discussed in the previos figure. The time window is set to 25% of overall period. This eliminates edges (and according nodes) from the graph that are too old and highlights the change of the network.

FIGURE5. Combined ego-network among the top 5 central authors.

node: author | link: co-authorship of adjacent authors | node color: top 5 most central nodes (egos) | node size: number of co-authorship contacts of author in visible combined ego-network | distance: strength of association between any two authors | node label: author-ID

Observation. The lighter color nodes in Figure 5 represent the top five central actors selected as ego. Interestingly, four further actors of the top6-top10 actors in Table 4 are also found in direct relationship to these top5. The peers of the top 5 ego nodes did not develop short cut links but are only linked indirectly via the ego nodes. This means, there is no strong effect of triadic closure.

FIGURE6. Ego-network among the top 5 central authors first vs. last years.

Observation. This figure separates the first half until 1991 and the second half of the sampling period. the development of the top 5 central nodes corresponds to the general development of the IRIS network and its shifting epicenters throughout the analyzed four time periods. Node 12 was very central during the first period. Later node 58 moved to the central position in period two. In this period, node 160 was embedded in a very small peripheral component, but it moved towards a very central position in period three (starting in 1993). The same applies for node 213, except that this author started its centralization efforts later in 1998. Node 343 finally is the most connected node in the fourth period although the first co-authorship link was established only in 1997. For the observed IRIS network, this implies that the active centers of the overall structure are formed by very active and prominent authors during their lifecycle of activity. Although three out of the top five have even collaborated directly, their respective groups of co-authors remain very separated.

FIGURE6a. Combined ego-network among the top 5 active authors.

Observation. Altogether, this subnetwork includes 84 authors in 198 relationships. The lighter nodes represents those top 5 central persons. Interestingly, only two of the top6-top10 active actors are found in direct relationship to the top5. The rest are positioned in different sectors of the network meaning that several independent collaborations arise.

FIGURE6b. Ego-network among the top 5 active authors first vs. last years.

Observation. Some connections between ego-networks were created. However, at other points the top5 actors disconnected. In both periods we found three (different) clusters and thus the connected network has never really existed but emerges because of the accumulative nature of SNA graphs (i.e. there is no decay). We can conclude that the most central nodes where establishing their prominent position largely independently of each other.

FIGURE7. The diffusion of the terms information and knowledge over time.

node: author | link: co-authorship of adjacent authors | node size: number of co-authorship contacts of author in visible combined ego-network | distance: strength of association between any two authors | node label: author-ID

Observation. The co-authorship network of authors, who employed the terms information and knowledge in their papers at three different points in time. This is indicating the spread of these terms over time. Note the different size of the network and its dispersion into many smaller subnetworks.

FIGURE 7a - MOVIE. The diffusion of the terms information over time
(5MB Quicktime .mov).

Observation. The same co-authorship network of authors, who employed the term information in their papers, indicating the spread of these terms over time. This time, the animation can be observed.

FIGURE 7b - MOVIE. The diffusion of the terms knowledge over time
(3MB Quicktime .mov).

Observation. The same co-authorship network of authors, who employed the term knowledge in their papers, indicating the spread of these terms over time. This time, the animation can be observed.

FIGURE 7c. Line Charts of keyword occurrence (trends) across time.

color: each of the top 20 keywords.

Observation. The absolute occurrence shows the constant occurrences of information, system, and development. Some terms like process, mobile, management, and knowledge increased in absolute occurrence. The percentual share of occurrence of those keywords (i.e. percent of papers with selected keyword versus overall papers) decreases noticeably over time. This highlights reduced coverage of the selected top twenty and indicates increased pluralism and heterogeneity: The few initial topics differentiate.

More related research can be found at: http://www.ikmresearch.de

*Numbering of Images corresponds to the Figure in the published paper. Additional material is marked with a subsequent letter, e.g. 3a)