Using Social Media to Represent 'Public Opinion'

The third presenter in this Future of Journalism 2017 session is Shannon McGregor, whose interest is in the role of social media in the construction of public opinion by the political press. There's an increasing tendency for journalistic coverage to claim that 'Twitter' or even 'the Internet' responded in a particular way to specific political issues and controversies, and social media certainly play a role in how public opinion is shaped, but how might we think about the type of public opinion that can be observed on social media?

We are able, of course, to measure aspects like retweets, @mentions, likes, shares, and other metrics, but this misses out on many other measures that are considerably more difficult to assess, especially where they exist in less public social media spaces. How public is such 'public opinion' on social media, in the first place? Such measurements are also instantaneous, and can change rapidly in line with changes in the live texts of political coverage.

Social media measurements are also unrepresentative, since social media userbases are not representative in the first place, and since the nature of political expression on social media in any one country is malleable and open to interference by domestic and foreign, human and non-human actors. Contrary to news stories that utilise conventional, well-designed political opinion polling, then, where such stories draw on the observation of opinions on social media are a great deal more problematic in what evidence exactly they are basing their discussion on – and those methodological details are rarely covered in detail in the stories themselves.

Shannon studied some 770 news stories about the 2016 U.S. election that related to social media, and followed up on this with 18 in-depth interviews with journalists. Practically none of these stories included any methodological detail; at best, they stated that the story was sparked by a trending hashtag on Twitter.

Journalists themselves have started to recognise this, and note that they are working with social media datasets as black boxes, since the datasets are proprietary and the methods used to create them – contrary to polling data – are not well described. Many journalists are simply relying on impressions from their own Twitter feeds to take the social media pulse, and they criticise their colleagues for doing so but readily engage in this approach themselves.

Where more quantitative data are being used, the quantifiability of social media data itself is fetishised as signalling some kind of reliability; this is also actively promoted by the social media platforms themselves, not least in election partnerships between Twitter or Facebook and news outlets.

Journalists therefore see social media as another tool alongside other election trends information, but some journalists also describe this as 'intellectually offensive'. But this adoption is also self-perpetuating; what journalists may be observing here is in part also their own ripple effect. By using social media data, systematically or impressionistically, to represent public opinion, journalists are validating such use even where there are significant internal concerns about these approaches are appropriate.