Social Networks That Help During Crises

Humans naturally exist in social networks where they exchange ideas and form relationships. Times of collective crises magnify the importance of our social embedding. Understanding how social networks are formed—and how they adapt to new circumstances—can help ensure policy responses are designed to slow the spread of bad things, like disease, while maximizing the spread of good things, like useful information.

Key to well-functioning social networks is collective intelligence, also known as the “wisdom of crowds.” Recent research in network science—the study of how interdependent individuals interact to form complex network structures—offers important insights on what makes a crowd “wise.”

To test the effect of collective decision making during an emergency, Shirado et al. conducted an experiment where they awarded payoffs to participants that depended on their ability to accurately predict if a disaster would occur and their subsequent choice to evacuate or stay. If there was no disaster, participants had higher payoffs if they stayed compared to evacuating on a false alarm. If there was a disaster, participants lost everything if they stayed. Participants communicated their predictions and choice to four “neighbors” in their network.

Along with a control condition of individual decision makers, participants were randomly assigned to two types of network structures: a random regular network and a ring-lattice network. These structures are graphs commonly used by network scientists to visualize social networks. In a random regular network, ties between any two individuals—the “nodes” of the social network—are formed randomly following probability distributions. The ring-lattice structure, on the other hand, is characterized by the transitivity of human relationships—something which is not present in random network models. When our friends are also friends with each other, and such relationships in the form of triangles are arranged to also know their neighboring triangles in a circular fashion, the resulting network resembles a ring-lattice structure. Compared to a random regular network, any two nodes are further apart in a ring-lattice network due to the high number redundant ties with closer neighbors. This means that the influence of information coming from a further source will be less immediate in a ring-lattice network compared to a regular network. Most real-world social networks are a mix of these two network types.

The researchers found that members of ring-lattice networks made decisions far more accurately than those of random regular networks. But ring-lattice networks were just as effective as individuals taking random guesses. This is because when information travels in social networks, it is assessed less for accuracy and more for its emotional signal. Due to normalcy bias—the tendency for people to assume a permanent status quo and underestimate the likelihood of a disaster—participants responded more to reassurances of safety than to disaster warnings, leading reassurances to spread faster than warnings and more participants not to evacuate when they should have. Just like engaging within a social network increases the risk of virus transmission, being part of a network increased participants’ risk of receiving inaccurate information.

While this sounds like bad news for the wisdom of crowds, Almaatouq et al. showed there are in fact conditions that make us stronger together. They hypothesized that social networks can make accurate decisions more often than individuals because of their ability to restructure and adapt to changing environments. This happens through two channels: plasticity and feedback. Plasticity describes how individuals can choose and re-choose who to follow in their network, making a network dynamic. Feedback describes post-decision information with which people can update beliefs and change behavior.

Almaatouq et al. tested this theory by conducting two experiments in which participants in different network conditions and a control group attempted to guess the correlation of a scatterplot. In the first experiment, plasticity was varied while feedback was held constant: both groups received feedback on their performance while only one dynamic group could re-choose up to three neighbors whom they consulted for each trial. The second experiment varied feedback while controlling plasticity: participants were assigned to groups with different levels of feedback on their performance, but all could re-choose their neighbors at each round. Finally, the authors compared their experiment results with a simulation of interacting agents who could update beliefs and rewire social connections.

Their hypothesis was confirmed: dynamic networks with full feedback made the most accurate predictions. While all network conditions outperformed the solo condition, dynamic networks made the best collective decisions by shifting influence to people with better information. The authors emphasize that dynamic groups with shifting internal connections only make better decisions if that dynamism also leads to higher-quality feedback.

Network research can also tell us what kind of social distancing is best to slow the spread of COVID-19 while minimizing the economic, social, and psychological costs of isolation. In their discussion of ideal social distancing strategies, Block et al. point out how disease can be spread through weak connections that form network bridges between tight communities. For example, when you interact with someone you don’t meet often, the likelihood of contracting a virus is higher than interacting with your immediate social circle because of the outsider’s access to unfamiliar groups. To limit COVID-19 transmission between groups, Block et. al propose behavioral strategies that could tighten communities and reduce the bridges between them that avail the spread of disease.

Through 40 simulations with 1000 actors, Block et. al tested three distancing strategies:

1) seek contact partners similar to oneself

2) make communities closer by limiting interaction to relationships embedded in triangles of contacts who are also in contact with each other, and

3) repeatedly interact with the same people to create bubbles.

They find that all three strategies significantly reduce disease transmission compared to the control condition of random contact reduction, which is similar to social distancing recommendations with no practical guidance. The third strategy that forms bubbles is slightly more effective than the two other strategies as tighter communities are formed with fewer bridges between them. Keeping interaction focused to smaller, stronger communities is the most promising social-distancing strategy that also preserves the benefits of social interaction.

Network science offers insights with immediate applicability to the coronavirus pandemic. To foster the spread of accurate and useful information, it is important to remember that social networks can be wiser than individual decision makers only if they can learn from their mistakes; they can also lead to inaccurate decisions when people falsely reassure each other. To curtail the spread of COVID-19, it is better to encourage tight-knit communities than ill-defined calls to self-quarantine so people can be healthy without being isolated. With this research, policymakers can harness social networks to help citizens weather crises of collective action.

---

Shirado, Hirokazu, Forrest W. Crawford, and Nicholas A. Christakis. 2020. “Collective Communication and Behaviour in Response to Uncertain ‘Danger’ in Network Experiments.” Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences 476, no. 2237. https://doi.org/10.1098/rspa.2019.0685.

Almaatouq, Abdullah, Alejandro Noriega-Campero, Abdulrahman Alotaibi, P. M. Krafft, Mehdi Moussaid, and Alex Pentland. 2020. “Adaptive Social Networks Promote the Wisdom of Crowds.” Proceedings of the National Academy of Sciences 117, (21): 11379–86. https://doi.org/10.1073/pnas.1917687117.

Block, Per, Marion Hoffman, Isabel J. Raabe, Jennifer Beam Dowd, Charles Rahal, Ridhi Kashyap, and Melinda C. Mills. 2020. “Social Network-Based Distancing Strategies to Flatten the COVID-19 Curve in a Post-Lockdown World.” Nature Human Behaviour: 1–9. https://doi.org/10.1038/s41562-020-0898-6.