Abstract
This work proposes a heterophily-based metric for quantifying polarization in social networks where multiple ideological, antagonistic communities coexist. This metric captures node-level polarization and is built on user’s affinity towards other communities rather than their own. Node-level values can then be aggregated at the community, network, or sub-network level, providing a more detailed map of polarization. We tested our metric on the Polblogs network, White Helmets Twitter interaction network with two communities and the VoterFraud2020 domain network with five communities. We also tested our metric on dK-random graphs to verify that it results in low polarization scores, as expected. Finally, we compared our metric with two widely used polarization measures: Guerra’s polarization index and RWC.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
References
Media Bias/Fact Check News. https://mediabiasfactcheck.com/. Accessed 25 July 2022
Abilov, A., Hua, Y., Matatov, H., Amir, O., Naaman, M.: Voterfraud 2020: a multi-modal dataset of election fraud claims on Twitter. In: ICWSM (2021)
Adamic, L.A., Glance, N.: The political blogosphere and the 2004 U.S. election: Divided they blog. In: Proceedings of the 3rd International Workshop on Link Discovery, LinkKDD 2005, pp. 36–43. Association for Computing Machinery, New York (2005)
Belcastro, L., Cantini, R., Marozzo, F., Talia, D., Trunfio, P.: Learning political polarization on social media using neural networks. IEEE Access 8, 47177–47187 (2020)
Borge-Holthoefer, J., Magdy, W., Darwish, K., Weber, I.: Content and network dynamics behind egyptian political polarization on Twitter. In: Proceedings of the 18th ACM Conference on Computer Supported Cooperative Work & Social Computing, pp. 700–711 (2015)
Conover, M., Ratkiewicz, J., Francisco, M., Gonçalves, B., Menczer, F., Flammini, A.: Political polarization on Twitter. In: Fifth International AAAI Conference on Weblogs and Social Media, January 2011
Darwish, K.: Quantifying polarization on twitter: the kavanaugh nomination. In: Weber, I., Darwish, K.M., Wagner, C., Zagheni, E., Nelson, L., Aref, S., Flöck, F. (eds.) SocInfo 2019. LNCS, vol. 11864, pp. 188–201. Springer, Cham (2019). https://doi.org/10.1007/978-3-030-34971-4_13
Demszky, D., Garg, N., Voigt, R., Zou, J.Y., Gentzkow, M., Shapiro, J.M., Jurafsky, D.: Analyzing polarization in social media: Method and application to tweets on 21 mass shootings. In: NAACL (2019)
DiMaggio, P., Evans, J., Bryson, B.: Have American’s social attitudes become more polarized? Am. J. Sociol. 102(3), 690–755 (1996)
Emamgholizadeh, H., Nourizade, M., Tajbakhsh, M.S., Hashminezhad, M., Esfahani, F.N.: A framework for quantifying controversy of social network debates using attributed networks: biased random walk (brw). Soc. Netw. Anal. Min. 10, 1–20 (2020)
Esteban, J., Ray, D.: On the Measurement of Polarization. Econometrica 62(4), 819–851 (1994)
Friedkin, N.E.: Horizons of observability and limits of informal control in organizations. Soc. Forces 62(1), 54–77 (1983)
Garimella, K., Morales, G., Gionis, A., Mathioudakis, M.: Quantifying controversy in social media. ACM Trans. Soc. Comput. 1, May 2015
Guerra, P.H.C., Meira, W., Cardie, C., Kleinberg, R.D.: A measure of polarization on social media networks based on community boundaries. In: ICWSM (2013)
Lozares, C., Verd, J.M., Cruz, I., Barranco, O.: Homophily and heterophily in personal networks. from mutual acquaintance to relationship intensity. Quality & Quantity 48, September 2014
Mahadevan, P., Krioukov, D., Fall, K., Vahdat, A.: Systematic topology analysis and generation using degree correlations. ACM SIGCOMM Computer Communication Review 36, June 2006
Morales, A.J., Borondo, J., Losada, J.C., Benito, R.M.: Measuring political polarization: Twitter shows the two sides of Venezuela. Chaos: Interdisciplinary J. Nonlinear Sci. 25(3), 033114 (2015)
Nair, S., Iamnitchi, A.: The polarized web of the voter fraud claims in the 2020 US presidential election. In: Workshop Proceedings of the 15th International AAAI Conference on Web and Social Media. International Workshop on Social Sensing (2021)
Nair, S., Ng, K., Iamnitchi, A., Skvoretz, J.: Diffusion of social conventions across polarized communities: an empirical study. Social Network Analysis and Mining 11, December 2021
Salloum, A., Chen, T.H.Y., Kivelä, M.: Separating polarization from noise: comparison and normalization of structural polarization measures. In: Proceedings of the ACM on Human-Computer Interaction 6(CSCW1), April 2022
Yang, M., Wen, X., Lin, Y.R., Deng, L.: Quantifying content polarization on Twitter. In: 2017 IEEE 3rd International Conference on Collaboration and Internet Computing (CIC), pp. 299–308 (2017)
Ortiz de Zarate, J., Di Giovanni, M., Feuerstein, E., Brambilla, M.: Measuring Controversy in Social Networks Through NLP, pp. 194–209. Springer International Publishing (09 2020)
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Appendix A Additional Materials
Appendix A Additional Materials
1.1 A.1 Edge Distribution of VoterFraud2020 Domain Network
Figure 1 depicts the distribution of edges to communities after relabelling unknown based on the dominant label in the node’s direct neighborhood. The majority of the edges in the right community are to themselves. Left-center also has most of its edges to left-center and right.
1.2 A.2 Visual Representation of Datasets
Polblogs. Figure 2a shows the visualization of Polblogs community structure. Light green represents liberal and red represents conservative. Figure 3a displays the Polblogs network colored based on the cross-community affinity of each node. The greater the value, the lighter the shade. Different nodes have distinct hues, which demonstrates that their values vary. The figure is dominated by a darker hue, indicating that the majority of nodes have low cross-community affinity values, resulting in a polarized network.
White Helmets Twitter Interaction Network. Figure 2b depicts the visual representation of communities in the WhiteHelmet interaction network. The colors red represents anti-White Helmets and green represents pro-White Helmets. Figure 3b displays the White Helmets network colored based on the cross-community affinity of each node. The greater the value, the lighter the shade. The figure is dominated by a darker hue, indicating a polarized network.
VoterFraud2020 Domain Network. The visual representation of communities in the network is shown in Fig. 2c. The color reflects the political orientation of the nodes, with red for right, orange for right-center, yellow for center, green for left-center, and blue for left. The right (47.4%) and the left-center (35.5%) constitute the majority of the network. The center has 7.4% nodes, the left has 6.1%, and the right-center has 3.6% nodes. Figure 3c shows the VoterFraud2020 domain network colored based on nodes’ cross-community affinity value. Darker hue means low cross-community affinity. Overall, the graph shows darker shade indicating that the network is polarization.
1.3 A.3 Scenarios of Network for CCA Calculation
Figure 4 depicts various scenarios of a network with seven nodes and two communities: red and green. For each scenario, the cross-community affinity for node v is provided. In scenario Fig. 4a all the immediate neighbors and two-hop neighbors of node v are members of the same community, indicating the absence of cross-community affinity. In this instance, CCA(v) has a value of -1.5. Similarly, in scenario Fig. 4f, all neighbors inside a two-hop neighborhood belong to the opposing community, resulting in a node with maximum cross-community affinity, where, CCA(v) equals 1.5. CCA(v) = 0 in the Fig. 4d, because the neighborhood of node v is equally distributed among both communities.
1.4 A.4 Ideological Distance for Five Communities
Table 3 shows the distance between communities in a scenario in which we consider \(C=5\). Right and left communities are on the ends of the spectrum. As a result, the distance between them is 1.
1.5 A.5 Relabelling VoterFraud2020 Domain Network
We randomized the labeling of the nodes of VoterFraud2020 domain network to see the effect on the polarization metric. We relabelled the network in three ways. In the first case, we randomly relabelled “unknown” without altering the community size of labeled nodes. We determined the number of “unknown” required by each community to maintain the community sizes. Then, “unknown” was arbitrarily assigned to each community. Even though we relabel “unknown” only, 75.6% of nodes in the network are “unknown”, making the network at least 75% random.
In the second case, we relabelled all the nodes in the network randomly but kept the number of nodes in each community the same as in the original VoterFraud2020 domain network. In the third case, the five labels are equally distributed to the network, thus creating five equal-sized communities. These random labeling are performed 10 times, and the results provided are averaged over these outcomes. The polarization scores for these experiments are given in Table 4. The left-center community score in the experiment where “unknown” labels are randomly assigned shows that the community is polarized, indicating that this network is not totally random. The network had negative polarization values in all randomization experiments, indicating a lack of polarization.
Rights and permissions
Copyright information
© 2022 The Author(s), under exclusive license to Springer Nature Switzerland AG
About this paper
Cite this paper
Nair, S., Iamnitchi, A. (2022). A Heterophily-Based Polarization Measure for Multi-community Networks. In: Hopfgartner, F., Jaidka, K., Mayr, P., Jose, J., Breitsohl, J. (eds) Social Informatics. SocInfo 2022. Lecture Notes in Computer Science, vol 13618. Springer, Cham. https://doi.org/10.1007/978-3-031-19097-1_32
Download citation
DOI: https://doi.org/10.1007/978-3-031-19097-1_32
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-031-19096-4
Online ISBN: 978-3-031-19097-1
eBook Packages: Computer ScienceComputer Science (R0)