Saturday, February 28, 2009

Same old, same old

Ethan Zuckerman argues here that homophilly -- the tendency for people to like people who are like them -- can lead to ignorance. This irritates me. Not because I disagree with it, in fact I think he has a point, just that it was something I wrote about in 2001 on a now-defunct website. The idea is that people will associate with people who share similar views and interets as them (among other things) and will also tend to read articles written by people who share their opinions. I suppose this is obvious. What is troubling about this tendency is that it is hard to explain from a psychological point of view. Of course there are explanations based on concepts such as 'identity' and 'self-esteem' but personally I find these unsatisfactory. To say, for example, that people associate with like-minded people in order to bolster their identity or to boost their self-esteem only rasises another question as to why our minds are designed with such fragile self-esteem or identity that it needs to be massaged by the present of similar opinion.

This is one of the reasons why I like evolutionary explanations, and I think the puzzle of homophily can be better answered by asking the question 'why would the mind be designed to want to associate with people who are like you?' How might this help us to leave behind more copies the genes that lead people to be homophilous (either directly, by having kids who survive to reproductive age or by our helping genetic reletatives in this regard).

The answer, I think, is quite simple and has two parts which are opposite sides of the same coin. The first is that associating with people like you reduces the possibility for conflict. People who have similar values and so on, are likely to want the world to be the same as you thus you are less likely to end up with conflicts relating to how the world should be. For example, people of the same political persuasion usually want to inhabit a similar world are are likely to work together to achieve that. People with different political beliefs want different worlds and this very fact can lead to conflict. The second, related, reason is that the more similar people are the more motivation one has to work in the interests of the other.

Of course, conflict can arise in even the most homophilous groups. This is because ultimately we have been designed to look after our own interests. (Before you draw breath to shout at your monitor of course I know that there are many examples of people acting altruistically and even laying down their lives for others, that is a very interesting subject that I will have to leave for another blog -- if I ever work out an answer). In The Selfish Gene Richard Dawkins makes an interesting observation regarding parasitism. He asks why selfish genes (usually) cooperate with one another when they find themselves in the same organism. Dawkins’s answer is that they cooperate because they all share the ‘interest’ of reproduction: all the genes of most organisms leave the host through the same exit point (the germ cells – sperm or eggs). Thus it is in each gene’s ‘interest’ that it cooperates with the others in maximizing its chances of achieving its goal of propagation; they are, in effect, all ‘playing for the same team’. If genes could propagate themselves by leaving the body by other, individual, routes then we would expect more conflict to occur. One reason why parasites are frequently harmful to the organism is that they often do not share the same exit point as the organism’s own genes.

Our own mitochonria (the power houses of the cells as biology teachers since time immemorial have called them) very probably started of as parasites -- they have their own private DNA, separate from "our" DNA which resides in the cellular nucleus. The theory goes that they gradually changed such that both mitochondial and cellular DNA now leave the body via the egg. At this point both share the same interest and 'cooperate' with one another.

So homophily has an upside: it potentially enables cooperation among genetically unrelated individuals. Evidence for this? So far it is weak (although someone might point me to relvant research). Research by Jens Binder and Andrew Howes at the university of Manchester suggests that the more diverse the friends on a persons social network site the more conflict the site owner reports. Perhaps more compellingly, in Marek Kohn's recent book on trust he cites evidedence that the most trusting societies tend to be those that are the most homogeneous. I hope it goes without saying that I report this as a research finding, rather than as a recommendation that we should attempt to socially engineer our societies by some system of ethic cleansing in order to increase societal trust: sometimes the cure is worse than the disease.

Homophily also has a downside which is the topic of Zuckerman's talk: it locks individuals into a system whereby they hear the same opinions over and over again to the point where whatever the opinion or value system happens to be to the extent that extreme and undesireably opinions can be normalised. In the eight-year old blog I mentioned above, I wrote of how the internet allows people with minority values in touch with one another. This is great if you happen to have a child who suffers from, say, Williams syndrome (a rare disorder affecting 1 in 10,000 live births) or depression. Such individuals can develop support networks and exchange tips and experiences with people whom they would be unlikely to bump into in the street. It also has a darker side allowing people with views that we might consider undesireable to meet, paedophiles for example (whether or not they happen to be radioactive as this hilarious 'news' item from the Daily Mail reports). Repeated exposure to such views can lead to such opinions being normalised.

The Internet may also, as Zuckerman argues in his talk and as my younger self argued in 2001, lead to individuals having a more restricted information diet. When reading a newspaper one's eye can be caught by articles one would not considered choosing to read. I concluded my article by suggesting that the Internet by reducing such serendipitous encounters can potentially narrow or experiences. But the internet has had enough negative press recently (see my previous blog posts on Susan Greenfield and Aric Sigman) and I do not wish to try and create a further moral panic (fat chance of that). I'm not even sure I believe it any more.

EDIT: (6/3/09) this paper (.PDF) suggests that if people are able to view the results of others in a problem solving exercise (analogous, I suppose, to using sites such as Delicious and Digg) then they tend to accept extant solutions rather than generate their own. Thus, they conclude, social bookmarking (etc.) sites can impeded creativity.


  1. I don't understand the empirical basis for homophily well enough. McPherson et al. (2001) does not help much, at least from my quick reading. Much of the work is so old that it is easy to believe that the statistical analyses were not very good.

    Also the rank order of similarities (race, gender, religion, education, status, and only then beliefs) suggests that homophily, as it is known is sociology, is a slightly different beast to what you mean?

    McPherson et al. also list "reasons" for homophily. The first reason is geography, and then kin.

    There does not seem to be much here to suggest that homophily is necessarily due to preferential selection by individuals? Rather, accidents of place and birth tend to a certain homogeneity in pairs?

    best, Andrew.

    title={{Birds of a feather: Homophily in social networks}},
    author={McPherson, M. and Smith-Lovin, L. and Cook, J.M.},
    journal={Annual review of sociology},
    publisher={Annual Reviews}

  2. You're quite correct race, religion, education are all come before beliefs. But I think that beliefs, interests, goals are the BIG thing and they naturally covary with race, religion, education (to some extent) and of course, religion (what IS religion if not beliefs, etc.). Some people describe two types of homophily, so-called baseline homophily which is to do with essentially your friends being forced upon you by propinquity (hate that word, but its what they use) and so-called 'inbreeding homophily' which is where people actually choose their friends. Most of the research IS rather old and I see SNS as being an interesting test of this the hypothesis as to whether it is accident (baseline) or choice (inbreeding) particularly in you look at online to offline friends (I have a study on this going through the interminable ethics process at the moment).

    There is a paper on this

    pretty crude as it looked at information expressed on myspace profile. This could lead to all sorts of problem cos it just looked at words which is a problem when you consider that someone who liked, say, "sisters of mercy" would not match with someone who like "the mission" even though they are both Goth bands, arguing for a technique using Latent Semantic Analysis or something. One of the findings is that gender no longer matters. Which is probably due to changes in society or something.

    Men and women's interests (as in life goals) are different (David Buss has done some research on the difference between same- and different-sex friendships) but generally they are more similar now that they used to be, hence the difference.

  3. Digg may freeze innovation among its users

    The author of this work is Rob Goldstone. He has done some interesting previous work that takes a cognitive science perspective on social networks. He is organising a symposium at this years Cog Sci conference in Amsterdam.