Posted on 12 November 2018 by Curtis Jessop, Research Director
For the last three years NatCen have hosted an event at Twitter UK as part of the ESRC’s Festival of Social Science, focusing on how social media data can be used in research for the benefit of society.
After I presented at last year’s event, I was excited to relax a little more this year. In and amongst enjoying Twitter’s hospitality and taking a few selfies with a Dalek, it was great to hear from and talk with so many social researchers engaged with what social media data can tell us about society. It was an opportunity to reflect on how things have changed over the past year.
The speakers were all very engaging. The diversity of the talks is a clear indication of how much is happening already and how much opportunity there is to do more in this field.
Kenn Cukier started by conceptualising the progress of social media research through dialectics. Initially heralded by its proponents (along with ‘big data’ more generally) as the solution to social research’s challenges of declining response rates and increasing fieldwork costs, it was also derided by its critics for its lack of representativeness. However, it has now developed into a credible methodology that recognises its limitations and that it works best when used alongside ‘traditional’ qualitative and quantitative approaches.
The speakers this year painted a picture of a mature methodology that better understands what it can do that existing methodologies can’t. Dr Cath Sleeman’s presentation covered a range of examples of using non-traditional data to provide insight into society. Her example of using data from over 40 million job adverts to identify skill shortages exemplified how the scale of big data can provide insight at a more granular level and more promptly than a traditional survey could. Of course, online job adverts aren’t a representative sample, and aren’t designed to answer research questions. This is why the data are most effective when used in conjunction with other sources.
Dr Elena Martellozzo’s research with the London Metropolitan Police used social media data to capture people’s behaviours in a ‘natural’ setting. Reflecting the breadth of applications of social media data, in contrast to the research using job adverts, Dr Martellozzo’s research took a more qualitative, ethnographic, approach – going beyond profile information to look at behaviour to better understand the grooming behaviours of sex offenders.
However, the experience of social media research over the past year hasn’t been all that rosy. Facebook’s experiences with Cambridge Analytica and the implementation of GDPR have made the public, social media platforms, and organisations conducting research more nervous about the use of social media data for research. What about consent? What about privacy? What about security?
Dr Steven McDermott talked about the use of social media data for research at the UK government department HM Revenue and Customs. It was an excellent example of the role this type of research can play in society. He has 400 researchers working across 50 projects in his department, which demonstrates the credibility social media research has developed as a methodology and its potential for direct impact on public policy. However, the nervous laughter from the audience that met his assertion that HMRC ‘are listening’ reflects the sensitivity of this topic and the importance of ethical discussions around social media research.
In the Q&A session following his presentation on using online data to measure (and predict) prejudice, Dr Walid Magdy reflected on the ethical problems of modelling social media data to infer characteristics in certain fields of research. Should researchers publish findings that could allow governments, or other actors, to identify particular groups in the population? What might be the implications for the human on the other side of the data when those models give a false positive?
In Joe Rice’s introduction from Twitter, he commented that Twitter data offer’s an excellent opportunity to understand the role of ‘influencers’ in society. Yet, much of its insight comes from the ‘mundane’. It is this banal day-to-day activity of people that I think is of most interest social researchers. It offers a bottom-up approach to data collection where people talk about what they want, in their own terms. However, it also reflects one of the key challenges it currently faces. Bots aside, behind every LinkedIn profile, Instagram picture and Tinder swipe is a person, and those people need to be assured that their data are being used responsibly or they will stop providing it.