This is the summary of an article by Marco Janssen. You can get the pdf of the behavioral targeting article here: Targeting Individuals to Catalyze Collective Action in Social Network.
A lot of the challenges that our society faces today are problems that require collective action, such as climate change. It is very important to study collective action within large heterogeneous groups. Governments can help provide collective action, but there’s more to the rules and laws implemented to cause a behavioral change. Some studies have shown that an individual has a greater chance in participating in a collective action through social influence, such as reputation and social pressure.
Is it possible to target the right, influential individuals so they can have a social influence on others leading to cooperative behavior? There have been studies related to spreading influence in the context of social networks and viral marketing and the primary role of peer pressure. This is the summary of a paper presenting a model of agents to solving a collective action problem, and describe which individuals to nudge.
The model is composed of agents that make decisions on which behavior to adopt, and each behavior corresponds to a certain personal reward (individual part). An agent’s action is also influenced by neighbors in the network (social influence part). The initial behavior is A, and the main question is what conditions are necessary for an agent to adopt behavior B. Individual utility is defined as how far away or close to the preference of an agent is to the behavior. Social influence, on the other hand, increases when there are more similar behavior around the neighborhood.
Model analysis is in the investigation of the effects of homophily, or the similarity of agent attributes in a certain network. Plus, interventions are also include, such a incentives which make a certain behavior more attractive than the already is, increasing the probability that an agent will make that decision.
The following are some of the model’s initial results. The default case used in model analysis is one wherein the agents can only obtain global feedback, which means that each agent has an equal share of the initial behavior. Investigation on the effects of local information is the next step, and it has been shown that adoption rate increases where homophily is at a high level.
Another batch of experiments investigates 4 types of interventions. It has been shown that those agents who are not socially influenced by the behavior of others are the ones more likely to be affected by these interventions. Those who have lots of connections are least affected because of the larger influence of peer pressure.
Improvement of this model can be done by including other human behavior assumptions including the process of deriving information and product preference changes, for example, due to changes in price. Testing the model on empirical data is also challenging.
Some of the interesting findings in this simplistic model include the following: the desired behavior can be increased to 400 percent if feedback on the adoption is given only to the agents in their social network, instead of on a global level. Second, the most effective strategy for increasing adoption rates is by targeting agents that are least likely to be affected by social influence in decision-making.