2.Social Resilience Research Based on Causal Discovery (2023.02–2023.05)

Social resilience refers to the remarkable ability of a social system to adapt, recover, and progress in the face of adversity, shocks, crises, or transformative changes. It embodies the capacity of a community or society to sustain stability, minimize losses, and swiftly restore normalcy when confronted with diverse challenges such as natural disasters, economic downturns, political upheavals, and technological advancements.

In the field of social sciences, social resilience stands as a classic research conundrum. Numerous factors influence social resilience, and various calculation indicators come into play. Nevertheless, most existing methods merely involve weighted calculations of multiple statistical indicators, lacking a widely accepted and efficient approach for calculating social resilience.

We try to analyze from the essence that people are the most basic elements of society, and the most important manifestation of social resilience should be through people’s feedback. In this study, we introduce a social cognitive network, a cognitive network operating at the societal level. Employing the causal discovery method, we uncover the causal relationships among diverse dimensions of social cognition to construct the social cognitive network. This dynamic network gathers data on psychological, cognitive, and behavioral aspects of society over a defined period. By gauging the network resilience of social cognitive networks, we calculate the evolving trends and characteristics of social resilience.

In contrast to other approaches that merely assign weights and calculate various indicators, the social cognitive network serves as a more comprehensive research tool for exploring the dynamics and inherent traits of social resilience. As a proxy for the social resilience system, it enables us to gain a profound understanding of how social resilience evolves over time.

For this study, we utilized four Reddit topic datasets, comprising posts and comments labeled under the topics of bitcoin, GME, ethereum, and covid. These datasets exhibit diverse scales, ranging from hundreds of thousands to tens of millions of data points.

Core Process

1.Build the social cognitive network:

Utilizing LIWC, we extracted valuable insights from Reddit posts and comments, representing the essence of each semantic dimension as the value for respective cognitive dimensions. Employing a daily time frame, we employed the causal discovery method to reveal the causal relationships between these various dimensions, creating a comprehensive cognitive causal map. Given that different causal discovery methods have varying scopes of application, we subjected multiple methods to rigorous testing, ultimately selecting the most stable approach for analyzing the global data. Below, we present an illustrative example of a social cognitive network:

2.Calculate the network resilience and effective state of the social cognitive network, and explore what kind of dynamic paradigm the social cognitive system belongs to:

In this section, we employ the GBB algorithm to condense the multi-dimensional structure of network resilience into a more manageable one-dimensional representation, facilitating analysis and description. Taking the Reddit_covid dataset as an example, intriguing patterns emerge as we observe the social cognitive network’s effective state diverging with the increase of the resilience coefficient. Consequently, the social cognitive system exhibits two distinctive states: one characterized by sustained high activity and the other maintaining a relatively lower level of activity. This remarkable bifurcation sheds light on the intricate dynamics of the social cognitive system under varying resilience conditions.

3.Changes in social resilience over time：

4.The manifestation of various cognitive dimensions when social resilience is low:

Observing the fluctuations in diverse cognitive dimensions and social resilience over time, it becomes apparent that a decline in people’s analytical thinking occurs when social resilience is low.

5.The impact of different disturbances on social resilience:

There are three disturbances to the social cognitive network, which are to delete nodes (eliminate a certain cognitive dimension), delete edges (eliminate the association between two cognitive dimensions), perturb the weights of all edges (the social environment has changed dramatically). By implementing the above three perturbations on the social cognitive network and observing the changes in the effective state, we found that deleting a certain cognitive dimension has the greatest impact on social resilience and is most likely to lead to social collapse.