Separating the truth from the buzz in social media
|
Version 0 of 1. There’s a growing buzz about social media. In fewer than 10 years it has risen to become a social activity of most people in the UK. More than half of Britons regularly use it, and across the world more than 50 social media platforms have in excess of 1 million users. Facebook is on course to have a larger digital flock than the population of any country. This has produced endless fascination about how social media is changing society: how it’s pulled the rug from under the feet of journalists; how it’s changing where political battles are fought; how it’s a new tool for terrorists and propagandists; and overall whether it’s making our lives richer or the opposite. But another, perhaps quieter, revolution has been happening too. In the spaces where research is carried out – at universities, in marketing companies, and across government – the rise in use of social media has been equally dramatic. Researchers now have to understand a society in which many people live at least part of their meaningful lives online. People use social media for everything from forging new social ties, sharing information about what is happening in society and what they’re doing, to talking about products, companies and their next prime minister. There is a sense shared by many researchers that the rise of social media has the potential to make their craft stronger, more powerful and more useful than ever before. The rise of social media has changed the complexion and speed of social data – information about people and society – that researchers can get their hands on. Thanks to social media, in the time it takes you to read this article, I could have downloaded more data-points produced by people than those gathered in the census every decade. Having recognised the scale of this opportunity, new analytics companies have sprung up measuring “reach”, “impressions”, “sentiment”, “share of voice”, and a host of other metrics, promising to tell you what’s happening on social media, what people are talking about, and (with a definite lean towards marketing the advertising), how your brand is faring against its competitors. Underlying this, the scale of social media data that feeds this kind of analysis has prompted the creation of new technologies and ways of analysing them, which are unfamiliar to social science. Overall, this new kind of research is very different from someone with a clipboard, and a statistician with a spreadsheet. There’s great opportunity, of course. Social science can suddenly make use of more data, richer data and more recent data than every before. But there is great danger, too. For over a century, social science has been set with the messy and difficult task of researching us, critically assessing – in everything from how the data is collected to how it is analysed and used – the pitfalls and problems that this endeavour entails. In a race for larger and larger numbers, more and more data, faster and faster, many of the most important principles that social science has painstakingly established have, generally speaking, been lost from social media research. Redressing this would be a very large catch: achieving rigour is the most important challenge facing social media research today. Unless social media research is transparent with its method, honest with its limitations and critical of its findings, it will not influence the kinds of important decisions across society that it ought. Demos and Ipsos Mori have been working on a project with Sussex University and CASM Consulting to bring together technologists and social scientists to work on this problem. This week we’ve launched a paper, The Road to Representivity, to argue for one of the most important principles of social science to also become an important principle of social media research: representivity. Representative research makes sure that what you’re measuring recognisably reflects a defined group – whether a given profession, type of voter, area of a country, or the whole of society. It isn’t always vital, and it might not always be possible, but in any research that wants to draw conclusions stretching beyond the people that have been directly researched themselves, it is crucial to understand how reflective your dataset is of your target population. In the case of social media research, it is just as important to ensure that findings are representative of the content in your dataset, than worrying only about how it compares to the offline world. Conducting representative research on social media is no easy task, we found we needed to peel back a number of different layers to the problem ranging across: Finally, there is the issue of comparability. Listening to something on social media is not the same as asking someone a question. Online spaces are public and active – they don’t just reflect society, they are also used by people to change it. This brings with it a body of new problems when you try to work out how online and offline research fit together. In the paper, we do our best to suggest how each of these problems can be confronted. Overall, it involves a mix of new technology we developed as part of the project different ways of analysing the data that you have, and thoughts on how it all should be understood, and how it connects with offline forms of research. We call this the “road to representivity”. It’s a new method that certainly isn’t perfect, but is an ongoing effort to put more sociology into big data (or equally more big data into social science). Overall, we have to remember what lies behind all this data: human beings. Social media research has to move beyond an obsession with ever bigger numbers, and become sensitive to our basic task – the messy, challenging task of understanding human beings through their everyday, lived experiences. Only then can we turn “buzz” into insight. Carl Miller is co-founder and research director of the centre for the analysis of social media at Demos. Steve Ginnis is the head of digital research of Ipsos Mori’s Social Research Institute. This work was funded by Innovate UK, the ESRC and EPSRC. |