A brief summary of my PhD research into designing better online and offline social networks.
[This is post one of three, looking at how important the shape of your social network is. The second looks at design strategies for social networks, and the third a the politics of social network data.]
The following is one of the most yawn-inducing statements in the universe: this is what I learned from my PhD. If you’ll give me just two seconds before you close the browser tab, I promise I’m going to climb out of my tiny PhD niche and do what you aren’t allowed to in your thesis — cut to the interesting bits. My research looked at how digital technology is making it possible to measure our social network in unprecedented detail, how digital technology is shaping our social connections, and how we can make design decisions with these facts in mind.
This post looks at some of the reasons you should care about social networks — by which I mean the network of social connections between people, whether digital or offline. Social connections powerfully shape our lives — happy or sad, fat or thin, even healthy or sick. I’ll also highlight the relatively low-quality data much historical social network research relies upon, and how things are starting to change. I’ll follow up with two more posts: a second post looking at research on how to design for better social networks; a third on the politics of social network data.
One of the discoveries of my PhD is a humbling sense of the tiny amount of research a single person can undertake — even over four years (4 years!). What I thought would be the bulk of my research — the practical work of building and testing social network analysis software — in fact only functions as the MacGuffin, a plot device to power me on my journey through the huge and dispersed existing literature on social networks.
I’ll spend the next paragraph outlining my practical work, then move on to the wider picture, which, despite everything (4 years!), I still find exciting.
I built and tested software (localnets.org) that analyses and visualises Twitter data to understand local community issues: people upset about when the bins are (not) collected, perceptions of crime, planning issues, etc. It also helped identify ‘community influencers’, civically active local individuals. Building a prototype served as a way into the wider political and ethical issues of how the public sector can use social media data.
My practical work was about social media data in local government, but I want to zoom out to talk about the importance of social connections in general. In short, people’s social interactions are fantastically important. We know this intuitively, we say that there is an “old boys network”, that it’s “who you know”, and many similar aphorisms. We understand that similar people form networks — “birds of a feather…”. In Chinese culture the term Guanxi reflects the same concept, referring to how well connected a particular individual is.
Digital technology makes it easier to measure social networks. It also modifies them. Perhaps the clearest example of technology modifying social connections is online dating — forming intimate connections between people who would never have met otherwise. Dating apps are fundamentally changing the way people meet their life partners, leading to increased marriages between ethnicities — that is, in the long term, changing demographics.
If dating apps reinforce a pattern where wealthy people tend to date another, and less wealthy people also disproportionately date each other, it will increase household inequality as rich people couple off and form double-wealthy households. If dating apps causes people to meet outside of their income-based social circles, it could reduce income inequality. Changes in who ends up marrying whom could easily contribute as much to income inequality as headline topics like executive pay. So far as I can tell, there is no research looking at income and online dating.
The history of social network research
It’s not new to observe that social networks are both important and easily altered. Influential psychologist Leon Festinger published research that captures both these ideas in 1950. 270 MIT students were randomly assigned to dorm rooms across 10 buildings. A follow-up survey asked the students who their three closest friends were. 65% came from the same building, 41% were in adjacent dorm rooms. Being even slightly physically closer to someone makes a social connection to them vastly more likely. This might be obvious, but the strength of the correlation is striking. Many people form their friendships for life at university. How many lives have been shaped by the accident of dorm room assignment?
In 1967, Stanley Milgram did an experiment to test the ‘6 degrees of separation’ concept, the idea that any random pair of humans can be linked by some small number of friends of friends. The only way to measure the average friend-of-friend path length in 1967 was to send out randomly addressed postcards, explaining the experiment and asking recipients to forward the letter to a designated target (name and approximate address were provided). In the (highly probable) case they didn’t know the target, the recipient was instructed to forward the letter to someone they knew who they thought would be more likely to know the target.
By examining the letters that arrived at the target address, Milgram could see that the postcard had been forwarded, on average, 5.7 times, and so concluded that there were, on average, 5.7 social connections between any two Americans. There are problems with this approach. Most letters got lost, decreasing statistical validity of the results, and people don’t know the optimal person to forward the postcard to in order to get it to its destination. The structure of our social connections is a fundamental fact about human society, yet, until the era of the web, it remained almost completely opaque. In 2012, research demonstrated very precisely that the equivalent to Milgram’s number in the context of Facebook users was 4.74 — and that’s across the whole world.
The idea of ‘social capital’ — roughly, the aggregate benefit you get from your social connections — goes back to the origin of sociology in the 1850s. We now have a wide range of evidence about the structure of our social connections and correlations with other phenomena. I wanted to highlight some of the most compelling results, but also to point out they are mostly based on weak data. We haven’t come that far since Milgram’s study. Even though the big tech companies almost certainly have exquisitely detailed data, making it available is highly problematic.
Contemporary social network research
The evidence we do have shows how important social connections are. Loneliness is often said to be a chronic issue in developed countries, and it is increasing according to some analyses. People who say they are lonely die younger and have worse mental health. Loneliness is, in some ways, the social network equivalent of poverty.
Social capital correlates with health and well being. Surveys also show that our subjective experience of happiness is anchored in our social connections: being richer doesn’t make you happy (past a certain point), but being richer than your friends does. Studies have also used this data to test for a link between social capital to GDP growth and found a small positive correlation.
Negative properties are transmitted by the network too. Being friends with obese people makes you more likely to be fat. If your friend becomes obese they can transmit obesity to you almost like an infection, as though their weight gain gave you social permission to follow suit.
But the empirical data around social networks is often weak. We only know about obesity ‘contagion’ because of the Framingham Heart Study, one of the largest medical studies ever conducted, running since 1948 and with over 5,000 participants. This kind of data is hard to come by.
Research mostly depends on proxy measures of social capital, for example ‘civic participation’ — membership of sports clubs or participation rates in national elections. Alternatively, surveys which ask questions like “Generally speaking, would you say that most people can be trusted, or that you can’t be too careful in dealing with people?” are assumed to reflect people’s social connections. (This question from the popular World Values Survey).
Sociometric badges are another way of gathering data about social networks. Sociometric badges are electronic devices worn on lanyards around the neck. They collect metrics such as speech patterns (through the microphone) and where and when those wearing them engage in face to face contact (through the IR transceiver). The badges can measure the offline, non-digital social networks in offices. Using this data researchers were able to identify high value social connections that were missing within the office network. Researchers then intervened, adding those social links by introducing people to one another. Their intervention reduced the number of emails sent and caused staff to describe their workplace as ‘better connected’. Alternatively, you can use the data to see how men and women form different networks, which could, in turn, inform interventions that might promote gender equality. These studies underline the benefits of high-quality data, and how new technologies are increasingly providing it. A sociometric badge might easily become a smartphone app.
Which brings us to the enormous privacy issues, the complex ethical terrain, and the political difficulties inherent in studying social networks. At the same time, it’s not a subject that can be ignored — the more technology allows us to understand network formation, the more responsibility we have. The more technology changes social networks, the more responsibility accrues to those making design designing.
Social network driven design
Sunstein and Thaler’s influential book Nudge also argues for those making design decisions to take ethical responsibility, but in the context of behaviour change. They open their book with a thought experiment about a school canteen: where should the fruit go, where should the chocolate bars go? Throughout their book they provide empirical evidence showing how powerful ‘choice architectures’ can be — placing the apples next to the checkout and hiding that chocolate bars can drastically change health outcomes. Once there is empirical evidence about fruit placement, responsibility ensues. There is no neutral setting: merchandising decisions are driving behaviour. Again, as we get an increasingly detailed picture of how social networks form, responsibility ensures.
There are, of course, risks, but there are also benefits. In my work on LocalNets, one unexpected discovery was the importance of local, non-chain shops in the social media ecology. A Costa coffee outlet won’t run it’s own Twitter account, but a local cafe often will, and when it does it often catalyses community social media activity around it. Twitter data gives us an empirical window on something we always suspected — Costa might be very efficient at pumping out coffee, but there is a social value to locally owned shops. The better the evidence for these social network structures, the stronger the argument against market forces which erode communities.
Sunstein and Thaler’s nudges have been applied to hypothetical school canteens, organ donation laws, making you pay your tax and providing clean drinking water in Morocco. Social network formation has a similar diversity of applications, from the algorithm behind your dating app to the spread of fake news, from increasing political polarisation to fostering innovation — even if the social network perspective is a more contested and less developed field than the mature nudge ‘industry’.
My next post will look at some design strategies that can help designers think about social network structures. A final post looks at the politics of social network data. I’ll leave you with Bill Withers telling us about the importance of social networks.