Maximizing the Spread of Stable Influence: Leveraging Norm-driven Moral-Motivation for Green Behavior Change in Networks

TL;DR

This paper extends a moral-motivation model for pro-environmental behavior to social networks, analyzing how local norms influence adoption and proposing optimization methods for maximizing long-term green behavior spread.

Contribution

It introduces a network-based decision model for pro-environmental behavior adoption, formulates optimization problems for influence maximization, and analyzes stability and computational complexity.

Findings

Network structure significantly affects behavior spread.

Integer programs can identify key nodes for influence.

Certain network classes exhibit different dynamics than complete graphs.

Abstract

In an effort to understand why individuals choose to participate in personally-expensive pro-environmental behaviors, environmental and behavioral economists have examined a moral-motivation model in which the decision to adopt a pro-environmental behavior depends on the society-wide market share of that behavior. An increasing body of practical research on adoption of pro-environmental behavior emphasizes the importance of encouragement from local social contacts and messaging about locally-embraced norms: we respond by extending the moral-motivation model to a social networks setting. We obtain a new decision rule: an individual adopts a pro-environmental behavior if he or she observes a certain threshold of adoption within their local social neighborhood. This gives rise to a concurrent update process which describes adoption of a pro-environmental behavior spreading through a network. The original moral-motivation model corresponds to the special case of our network version in a complete graph. By improving convergence results, we formulate modest-size Integer Programs that accurately (but not efficiently) find minimum-size sets of nodes that convert the entire network, or alternately that maximize long-term adoption in the network given a limited number of nodes which may be temporarily converted. Issues of stability in determining long-term adoption are key. We give hardness of approximation results for these optimization problems. We demonstrate that there exist classes of networks which qualitatively have severely different behavior than the non-networked version, and provide preliminary computational results in in modestly-sized highly-clustered small-world networks related to the famous small-world networks of Watts and Strogatz.