I had a little dialogue with Meredith Rolfe after reading her papers on political participation and social networks:

AG:
You write, “why would a rational individual participate in an election in which the chance of her vote influencing the outcome of the election is essentially zero?”

Actually, there is a good reason. Basically, the evidence all points to the idea that people vote for what they perceive is beneficial to the country–and the stakes can be high enough that it is rational to make that vote. The ideas of “social preference” in our paper seem related to your ideas of social networks (although you are working at a more sophisticated level than we are).

Just thought you might be interested in these connections–also, I didn’t want you to dismiss the rational explanation so quickly. As we discuss in our paper, “rationality” is often (and unnecessarily) conflated with “selfishness”.

MR:
I do start (at least in my own mind!) with the assumption that voter turnout is largely a product of the perception that turnout is good for the all of the citizens in a population (like you, I try to avoid the functionalist fallacy by stressing the perceived benefits of voting, instead of the actual benefits.)

In my work, though, I focus not on the social preferences themselves, but on conditional compliance with this widespread belief that voting is good for the country. I assume that about 10-15% of the population is willing to contribute unconditionally, or vote regardless of what the people around them are doing (these are, in essence, the people who may simply weigh up social benefits and personal costs and cooperate). The vast majority of citizens, though, are simply responding to the decisions of those around them, with the importance of the election being a primary determinant of how many people they will discuss politics with. In other words, their decisions are conditional (either linear/ “fair”, or non-linear/ “conformist”.)

One question I had was how a social preferences model would deal with the fact that people can acknowledge that something is good for the group, and still not do it (most non-voters claim that voting is always the right thing to do, and most players in social dilemmas claim that it is best to contribute at least half of a personal endowment to the group).

Also, in an experimental situations, it is possible to get a person to change the amount they are willing to contribute to the group without changing the benefit to the group (N or the $ amount/importance of election), but by changing their perceptions of who will find out about their decision, or the amount that they expect/observe to be given by Others?

Do you think it is possible to generate conditional decision-making from a dual utility model? I don’t believe that conditional cooperators are irrational, either in the sense of ordered preferences, or, I suspect, material self-interest as well. However, it is less clear to me in the short term what the logical or analytical path would be between social benefits and conditional responses in local (versus global) social networks. That is why I chose to dock the model to empirical evidence instead of a model of rational preferences.

In the next few years, I have in the back of my mind playing around with either an axelrod-esque competition where groups of strategies have to compete with each other (funny that the latest round of his competition was actually won by a group of strategies!) or allowing single players to play n-player games within social networks or maybe even trying to see how far I could push the notion of correlated equilbria. I’m hoping that these efforts would produce an adaptively rational mooring for the conditional cooperation model, but who knows…

On to the topic of network models:

MR:
Are you familiar with some of the p*/exponential random graph models? These models focus on the ways that triads (not dyads) deviate from random expectations, but still with some eye to looking at the role of homophily/individual characteristics in friendship formation. I’d be interested in your observations of how it relates to the approach you propose. Tom Snijders is the guy to check out if you aren’t familiar with this relatively new body of work.

AG:
I wasn’t familiar with the random graph models you mention. I think that for problems such as social/political polarization, the issue is not so much the structure of the graph but rather how certain “labels” (e.g., michaels, or american indians, or liberals or conservatives) are distributed in that network.

I took a look at your paper on conditional decision making. It’s an intersting topic and I like the idea of it being considered from a more political-sciencey perspective. My big idea here is that the news media should be considered as part of the network. It’s not all about people talking with their friends.

MR:
You’ll see what you think about the ERG models in terms of the underlying statistical assumptions. Technically, you can use them to estimate the effects of categorical variables on the way friendships are distributed. What I like about your approach is that it is more suitable to use with survey data (or the Killworth/Bernard/McCarthy populations type data) instead of hard to collect data on complete populations. I think that is really important for political scientists who rely heavily on survey data.

I’m pretty torn on the issue of the media. It is obviously an issue. I saw a recent presentation by Jamie Druckman that was pretty strong in the claims for influence of media on candidate choices and descriptions that really raised some questions for me. Plus, my husband is also a media guy, he looks at how media imagery affects racial attitudes. But, what do you do with the fact that media coverage is designed to attract viewers? That politicians take stands to be popular? But you know, you may be very right, by putting media into the “network” it could both be influential and be influenced by the people it was connected to, hmmm… That is actually a very, very cool idea. I’ll think about it more.