5.3.1 - RQ1: What is a simple model that can serve as a proxy for behaviors that indicate Group Polarization?
Figure 19 - RQ1 Components
The goal of building a computer model in this case is to have a reference to inform the design process, and later, as data is gathered from users, to be updated and refined by real-world results. The use of such models based on simplifying assumptions has been used in fields as widespread as climate and economics. For social systems that are based on the interaction between individuals, agent-based approaches are widely used. The use of agent-based computer models for such ‘generative social science’ is appropriate for building systems where complex behavior emerges at the result of interaction of simple rules with and environment. In many respects it is the direct descendent of Simon’s Ant. Joshua Epstein provides a thorough overview of such systems in Agent-Based Computational Models and Generative Social Science (Epstein, 2006, p. 5 - 10).
Eric Bonabeau, in a colloquium on agent-based simulation describes five situations where agent-based simulation is appropriate. The model in this proposal aligns with the social network situation, where social networks are “are characterized by clusters, leading to deviations from the average behavior” (Bonabeau, 2002, p. 7287). Macal and North state that Agent-Based Modelling and Simulation (ABMS) is “a preferred mechanism to represent social interaction, collaboration, group behavior, and the emergence of higher order social structure” (Macal & North, 2005., p. 3), The social characteristics of Group Polarization as supported by information bubbles and anti-bubbles fall into this situation.
Group Polarization is a social phenomenon,that emerges from individual behaviors. Work modelling of the social aspect of this has been done by Guillaume Deffuant, who explored the creation of extreme, polarized, opinions (Deffuant, Amblard, & Weisbuch, 2004). Yu and Dayan that developed a neurological behavior model for explore/exploit behavior in individuals (Yu & Dayan, 2005).
The core agent-based simulation for this effort will begin with the following components and be refined from this initial point as user data becomes available:
- Fundamental bias - Each agent will have a type of EXPLORER, CONFIRMER, AVOIDER, or SOURCE. These types will behave differently with respect to new information.SOURCE may be different from the others, since SOURCE agents will original information to the other agents, and may behave according to external experimental constraints.
- Information - We need a small piece of information that can be used by these simple agents, but still manifest in such a way that individual or group patterns. I propose using a simple model consisting of ‘statements’ and ‘beliefs’.
- A statement is a small class that contains a name, weight, credibility and accuracy. These are arbitrary values, that are created by SOURCE agents with values for all but weight, and then used by other agents. Upon incorporation, the agent sets an initial weight. This is the value that the agent places on its copy of the statement.
- A belief is a set of statements. Agent beliefs can be compared with each other. Comparing alters the weight of some statements. Agents with similar beliefs are attracted to each other, while agents with differing beliefs are repelled. These ‘forces’ can be applied in n dimensions resulting in relative positioning of agents. Implicit in this design is the ability to use distance as a proxy for similarity that can be used for a variety of useful calculations, such as cluster calculations. It also allows for visualization, which should be extremely helpful.
- SOURCEs produce their statements and beliefs according to the rules of the experiment. These SOURCE values might change over time, and converge or diverge.
In the basic system that may be sufficient to see effects. The rules for individual agents would be something along the following lines:
- Initial Belief formation - An agent randomly samples a set of SOURCEs and draws a set of statements from the SOURCE beliefs.
- Iteration - At each subsequent turn, each agent interacts with their neighboring agents such that the differences in behavior are expressed. EXPLORERS should sample more widely, CONFIRMERS should look for matching beliefs/statements, while AVOIDERS should look only at agents who already share similar beliefs. An example of a potential selection algorithm is described by Lande, who creates a model for runaway sexual selection as originally described by Darwin and Fisher (Lande, 1981). Runaway selection is an instance of a confirming pattern and aspects of Lande’s model will be included in the model.
220.127.116.11 - Evaluation Criteria:
A successful model should display two qualities:
- The differences between EXPLORER, CONFIRMER and AVOIDER should be statistically significant. Within the model, this should mean that the quantitative data about agent sources should be distinguishable. For example, EXPLORERS should have a wider range of sources. This should manifest itself as a greater variance. CONFIRMERS should have tighter variance, with a higher total of sources than AVOIDERS. This should be calculable using bootstrapping, using Chapman and Hall’s An Introduction to the Bootstrap as my primary authority (Efron & Tibshirani, 1994). I would expect a 95% confidence between EXPLORERS and others. I’m not sure that there will be as large a difference between CONFIRMERS and AVOIDERS. To a degree this may be moot, as the simulation can be run over sufficiently large enough populations to empirically determine the needed statistical power.
- Group polarization should emerge during simulation, and the degree of polarization should depend on the distribution of agents. To confirm the creation of bubbles, I propose the use of dp-means or DBSCAN cluster detection. The repeatable creation of unique clusters centered around shared ‘beliefs’ should indicate the presence of group polarization. Again, this should be confirmable by comparing the members of a cluster to each of the other clusters using bootstrapping analysis of means and variances. My intuition is that there will be considerable ‘tuning’ of parameters to achieve a statistically meaningful separation of an unknown number of clusters. My hope is that these parameters will provide insight into the analysis of human interaction in finding answers to the following research questions. This type of interaction between model and ‘real world’ data has been explored using cell phone data by (Roehner, 2005) though I am having difficulty finding other examples.
 This algorithm implements Fisher's solution of Darwin's problem of why in many species with polygamous systems of mating females should prefer mates with extreme characters that are apparently useless or deleterious for survival, such as the plumage of some male birds and the horns and tusks of certain male mammals.
Fisher showed that a positive correlation between female mating preferences and male secondary sexual characters will arise in the population because genetic variance in the preferences of females for more extreme males. The evolution of mating preferences may be self-reinforcing because once started, females are selecting not only for more extreme males but also indirectly, through the genetic correlation, for a higher intensity of mating preferences.
Fisher described this as a runaway process that must eventually be stopped by severe counterselection against extreme males or against the most discriminating females because of their difficulty in finding a suitable mate.