Chapter 23: Further Reading — Network Analysis of Information Spread

Annotations describe each source's contribution and relevance to the chapter's themes. Sources are grouped thematically.

Foundational Network Science

Watts, D. J., & Strogatz, S. H. (1998). Collective dynamics of 'small-world' networks. Nature, 393(6684), 440–442.

The founding paper of small-world network theory. In six pages, Watts and Strogatz introduce the mathematical framework for characterizing networks that combine high clustering with short path lengths. They demonstrate these properties in three real-world networks (neural, power grid, actor collaboration) and provide the generative model (random rewiring of a regular lattice) that produces small-world properties. Essential reading for understanding why social networks are such efficient information propagation systems. The paper is technically accessible without advanced mathematics.

Barabási, A.-L., & Albert, R. (1999). Emergence of scaling in random networks. Science, 286(5439), 509–512.

The foundational paper on scale-free networks. Barabási and Albert show that the power-law degree distributions observed in the World Wide Web and actor collaboration networks arise from two mechanisms: network growth (new nodes are continually added) and preferential attachment (new nodes connect to existing nodes in proportion to degree). This paper is the starting point for understanding why social media platforms have a small number of extremely influential accounts and a long tail of low-influence accounts.

Barabási, A.-L. (2016). Network Science. Cambridge University Press. (Available free online at networksciencebook.com)

The definitive textbook on network science, available online at no cost. Covers everything in this chapter and much more, with beautiful visualizations and interactive elements. Chapters 4 (Scale-Free Property), 5 (The Barabási-Albert Model), 9 (Communities), and 10 (Spreading Phenomena) are directly relevant. Suitable for readers who want to go deeper into the mathematics without requiring graduate-level probability theory.

Watts, D. J. (2003). Six Degrees: The Science of a Connected Age. Norton.

Watts's accessible account of the science of networks, written for a general audience. Covers the historical development of network science from Milgram's small-world experiment through to the mathematical formalization of the 1990s. Particularly valuable for contextualizing network science within broader social science traditions. Watts also engages critically with oversimplified "six degrees of separation" narratives that overstate the implications of small-world structure for individual-level reachability.

Empirical Misinformation Spread

Vosoughi, S., Roy, D., & Aral, S. (2018). The spread of true and false news online. Science, 359(6380), 1146–1151.

The landmark empirical study analyzed in depth in Section 23.5. Analyzes 126,000 stories tweeted by 3 million people from 2006–2017, finding that false news spreads faster, farther, and deeper than true news, driven primarily by human behavior rather than bots. The paper's supplementary materials contain extensive methodological detail and robustness checks. Anyone working in misinformation research should read the full paper, including the supplementary information.

Goel, S., Anderson, A., Hofman, J., & Watts, D. J. (2016). The structural virality of online diffusion. Management Science, 62(1), 180–196.

Introduces the concept of structural virality — a metric that characterizes whether information spreads through broadcast (a single hub reaching many directly) or viral (long chains of one-to-one sharing) mechanisms. The paper analyzes millions of Twitter cascades and shows that truly viral spread (in the technical sense) is far rarer than commonly assumed. Essential companion to Vosoughi et al. for understanding cascade structure. Provides the mathematical definition of structural virality used in Case Study 23.2.

Altay, S., de Araujo, E., & Acerbi, A. (2022). Quantifying the relation between trustworthiness and spreading of COVID-19 misinformation. PLOS ONE, 17(1), e0261257.

An important qualification of Vosoughi et al.'s findings. Altay et al. show that while the most-shared false news items do spread more than the most-shared true news items (replicating Vosoughi et al.'s finding for the viral tail of the distribution), the average false news item circulates very little. Most misinformation remains obscure. This finding suggests that the "fastest and farthest" framing may overstate the prevalence of harmful false content while underweighting the few but highly viral items that cause most damage. Crucial for avoiding moral panics about the scale of the misinformation problem.

Juul, J. L., & Ugander, J. (2021). Comparing information diffusion mechanisms by matching on cascade size. Proceedings of the National Academy of Sciences, 118(35), e2100786118.

A methodological critique and extension of Vosoughi et al. (2018). Juul and Ugander argue that comparing cascades of different sizes (which are prevalent in the false vs. true news comparison) can produce spurious structural differences. They propose a matching approach that compares cascades of equal size and find that some (but not all) of Vosoughi et al.'s findings are robust to this correction. Essential reading for anyone citing or building on the Vosoughi et al. findings.

Community Detection and Echo Chambers

Blondel, V. D., Guillaume, J.-L., Lambiotte, R., & Lefebvre, E. (2008). Fast unfolding of communities in large networks. Journal of Statistical Mechanics: Theory and Experiment, 2008(10), P10008.

The original Louvain algorithm paper. Describes the two-phase modularity optimization procedure and demonstrates its application to networks with up to 120 million nodes. The algorithm described here is what is implemented in the python-louvain package used in this chapter's code examples. The paper is concise and technically accessible.

Newman, M. E. J., & Girvan, M. (2004). Finding and evaluating community structure in networks. Physical Review E, 69(2), 026113.

Introduces both the Girvan-Newman edge-betweenness community detection algorithm and the modularity metric Q. Newman and Girvan's contribution of the modularity metric is arguably more influential than their algorithm — modularity is now the standard measure for evaluating community structure quality. The paper provides the mathematical definition of modularity used throughout this chapter.

Bail, C. A., Argyle, L. P., Brown, T. W., Bumpus, J. P., Chen, H., Hunzaker, M. B. F., ... & Volfovsky, A. (2018). Exposure to opposing views on social media can increase political polarization. Proceedings of the National Academy of Sciences, 115(37), 9216–9221.

A counterintuitive experimental finding: exposing Twitter users to cross-cutting political content (content from opposing political viewpoints) actually increased their political polarization rather than reducing it. This challenges simple narratives about echo chambers (that exposure to diverse views automatically reduces polarization) and suggests that the relationship between network structure, information exposure, and attitude formation is more complex than assumed. Essential for nuanced discussion of echo chamber effects.

Influence Maximization and Diffusion Theory

Kempe, D., Kleinberg, J., & Tardos, É. (2003). Maximizing the spread of influence through a social network. Proceedings of the 9th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 137–146.

The foundational paper on influence maximization. Kempe, Kleinberg, and Tardos formally define the IC and LT models, prove that influence maximization is NP-hard, and prove that the greedy algorithm achieves a (1 - 1/e) approximation guarantee due to the submodularity of influence functions. This paper is the theoretical backbone of Section 23.7. The mathematical proofs are sophisticated but the conceptual intuitions are explained clearly.

Centola, D., & Macy, M. (2007). Complex contagions and the weakness of long ties. American Journal of Sociology, 113(3), 702–734.

Introduces the distinction between simple and complex contagion in network diffusion. Simple contagion (like disease or viral content) spreads efficiently through weak ties (bridges between communities). Complex contagion (behaviors requiring social reinforcement) spreads more effectively through strong ties within dense clusters. This distinction is critical for choosing between the IC model (appropriate for simple contagion) and the LT model (appropriate for complex contagion). The paper also challenges Granovetter's "strength of weak ties" finding for complex behaviors.

State-Sponsored Information Operations and Platform Data

Howard, P. N., Ganesh, B., Liotsiou, D., Kelly, J., & François, C. (2018). The IRA, social media and political polarization in the United States, 2012-2018. Oxford Internet Institute Computational Propaganda Project.

The most comprehensive analysis of the Twitter IRA dataset, produced by the Oxford Internet Institute for the US Senate Select Committee on Intelligence. Covers content analysis, network analysis, and temporal analysis of the IRA's Twitter operation. Provides the empirical foundation for Case Study 23.1. Freely available online. Note that some conclusions about the IRA's impact on political polarization have been contested in subsequent research.

Benkler, Y., Faris, R., & Roberts, H. (2018). Network Propaganda: Manipulation, Disinformation, and Radicalization in American Politics. Oxford University Press.

A comprehensive network-based analysis of the American political media ecosystem using hyperlink networks, social media data, and traditional media content analysis. Benkler et al. argue that the primary driver of misinformation in US political discourse is not Russian interference or bot activity but asymmetric polarization in domestic media — specifically the transformation of right-wing media into a self-reinforcing network that operates by different epistemic norms from mainstream media. A landmark work for understanding the domestic network structure of political misinformation. The data and methodology are described in detail and the book is available open access.