We understand the dynamics of the world around us as by associating pairs of events, where one event has some influence on the other. These pairs of events can be aggregated into a web of memories representing our understanding of an episode of history. The events and the associations between them need not be directly experienced—they can also be acquired by communication. In this paper we take a network approach to study the dynamics of memories of history.
First we investigate the network structure of a data set consisting of reported events by several individuals and how associations connect them. We focus our measurement on degree distributions, degree correlations, cycles (which represent inconsistencies as they would break the time ordering) and community structure. We proceed to model effects of communication using an agent-based model. We investigate the conditions for the memory webs of different individuals to converge to collective memories, how groups where the individuals have similar memories (but different from other groups) can form.
Our work outlines how the cognitive representation of memories and social structure can co-evolve as a contagious process. We generate some testable hypotheses including that the number of groups is limited as a function of the total population size.
Figure 1. Structural properties of an empirical aggregated memory web.
(A) displays cumulative in- and outdegree distributions. (B) shows the clustering coefficient as a function of the in and out degrees. In panel (C) we investigate the degree–degree correlation by plotting the average neighbor degree as a function of the in- and outdegree of an agent. (D) is a plot of the largest connected component of the aggregated memory web, highlighting the modular structure. A weighted-network clustering scheme identifies 48 smaller groups. If these are treated as vertices and the same clustering scheme is applied to that network of groups, we discover four supergroups indicated by different colors.
Our model has three basic assumptions. First, that people, while communicating, influence each other's associations; so that the higher the frequency with which an individual hears another person make an association between events, the stronger that association becomes in the individual's memory web. Second, the stronger an association is, the more likely an individual is to talk about it. Third, people are more likely to communicate with people that they perceive as similar to them (in terms of age, interests, location, etc
Figure 2. Illustrations of the model.
(A) shows an illustration of the social network of agents. The thickness of the lines are proportional to the strength of the social connection—the familiarity F. The social networks can have communities (encircled in the figure)—groups of agents that are more strongly connected
The concept of memory has both psychological and social dimensions In historical discourse, it has lately drifted more towards the latter . We model collective memory based on five principles—the experience of actual events, communication across social networks, reinforcement of both memories and social ties from the communication, errors and misconceptions, and forgetting. Our model takes an external memory web as input, a network meant to represent a hypothetical web of associations of unbiased, well informed but otherwise normal individuals. We use an empirical dataset for this seed network constructed from the life-stories of fourteen Chinese villagers. This dataset says something about memory webs in its own right. Most of the events are connected into a large component—the villagers can associate one event to another and thereby cover most of the important events around them during their lives.The model predict three different scenarios of collective memories—either the memories are very personal (each agent has its own view of history), or all agents think more or less the same, or (the intermediate case) there are distinct groups which share the same view of history.