What Big Data Can(not) Tell us About How Twitter Is Used

Lori McCay-Peet blogs about her BD&S article with Anabel Quan-Haase and Kim Martin, ‘Networks of digital humanities scholars: The informational and social uses and gratifications of Twitter'

by Lori McCay-Peet

Social media are often associated with Big Data because of the massive amounts of data that are generated through activity on sites such as Twitter and Facebook by people, businesses, organizations, and governments. With Big Data, big, broad questions about an entire social media population can be posed as well as questions about the nature of the platform itself.

For example, through Big Data analytics, it’s possible to understand what Twitter is. Is Twitter a social network, enabling users to connect and communicate with one another? Or is Twitter an information network, allowing users to broadcast and gather information? The answer, boiled down, is simply ‘both’—though couched with ‘it depends.’ A more nuanced understanding of what Twitter is can be difficult to discern without peering closely into the clusters of networks that have developed on Twitter to see what motivates individual users to adopt Twitter as a social media platform and probe why and how they use it. In other words, it’s important to take social context into consideration.

In our paper, we examine the uses and gratifications of Twitter in the context of the scholarly practice of digital humanities (DH) scholars. Rather than Big Data, we approached the question of the uses of Twitter through interviews with DH scholars. In the process of answering the social versus information network question, we discovered some of the factors that compel DH scholars to use Twitter even when it’s noisy, near impossible to fit their ideas into 140 characters, and hard to make the time to do it. We found that while invisible colleges, informal communication networks of specific research areas, appear to be alive and well on Twitter and DH scholars have a particular affinity for Twitter in this respect, their use of Twitter is complex and varies by scholar and point in time. For example, one DH scholar we interviewed used Twitter to follow people with expertise in a particular area because “I needed to know more, I needed to ‘skill up’ quickly,” suggesting the timely, informational value of Twitter. And while Twitter is undoubtedly an information network, used for dissemination or maintaining awareness of ideas and information, it is also a social network. The people we interviewed discussed the personal exchanges they experienced among DH scholars and they expressed an awareness of who was following them on Twitter which highlights Twitter’s use as a social network.

While more small study research is needed to delve into the intricacies of DH scholars’ and other groups’ social media use, how can we now hand the question of information versus social network back over to the Big Data researchers? In our paper we echo the call by Brooke Foucault Welles (See ‘On minorities and outliers: The case for making Big Data small’) to make Big Data small, to examine subsets of social media populations to understand outliers, minorities, or in the case of DH scholars, invisible colleges made visible on Twitter. How does the DH subset of the Twitter population compare, for example, with other subsets when examining indicators used in Big Data analytics (e.g., degree assortativity, path length between users)? We argue small study research findings may help explain Big Data results and vice versus. What other research paths can we find and develop when we think to pair small study research with Big Data approaches?

About the author

Lori McCay-Peet is a postdoctoral fellow in the Department of Sociology at the University of Western Ontario. Her research examines people’s perceptions and uses of web-based technologies.

Twitter and Censorship: Can Images and Big Data Help Hack the Silence?

by Paolo Cardullo

In this post, I would like to reflect further on the process of gathering and analysing digital data from Twitter. In particular, I want to think about imaginative ways of working with digital images which immensely circulate on social media.

When I first started observing the Turkish censorship of Twitter in March 2014, I immediately thought that the sole use of visually powerful infographics and traffic analysis did not effectively explain what was going on, both online and behind the scenes of the 'digital coup'. I have been for a long time an Open Source software and digital rights activist, as well as a trainee hacker. I therefore found it hard to believe that data was seamlessly flowing despite attempts from the Turkish government to block the popular social platform. I immediately understood that graphics and tables were hiding a lot more than they were saying. I therefore started a process of intense following, outreaching a few active participants for online interviews, as well as collecting most reflexive tweets and images. The full story is narrated in my article for Big Data & Society. Here, I would like to show how images, triangulated with other sets of data, reinforce the theoretical arguments I make in the paper: namely, that during systemic chock points, such as censorship of the Internet, ordinary users enact a series of unpredictable tactics of circumvention of censorship, aided by more traditional hackers. Drawing from Italian Autonomists' ideas around digital labour and acknowledging the learning process attached to use of ubiquitous digital devices, I put together the provocative concept of a 'hacking multitude'. The paper suggests that a generalisation of hacking/multitude can be very problematic for Big Data analysis, because of the unacknowledged potential that a multitude has for modifying, hiding, or pushing through the flow of digital data.

In this ensemble of different data point, a special place belongs to images. They circulate immensely on digital platforms – they are increasingly 'poor images' (Hito Steyerl, 2009), highly compressed in order to circulate faster. Despite this, visual material is hard to take into account in traffic and metric analyses. Images are Big Data too, of course: a combination of bits and pixels that escapes algorithm identification or categorisation. Images in fact need to be contextualised, explained, and analysed (also in relation to other visual discourses). In the article, I broadly discuss one particular image (see image above) which triggered my intuitions on what a 'hacking multitude' might be. It is this image that somehow determined in me, the researcher,  a Barthesian 'punctum' around transformative moments. I would contend that more traditional Big Data representations (metrics, infographics, tags, etc) can hardly reveal such a textured understanding of the social world.

About the author

Paolo Cardullo is an Associate Lecturer in Sociology at Goldsmiths College, University of London, and a visiting research fellow at the Centre for Urban and Community Research (CUCR). He finished his PhD in Visual Sociology at Goldsmiths (UoL), with a thesis around the affective geographies of gentrification in East Greenwich (2012). ‘Walking on the Rim: Towards a Geography of Resentment’ was discussed with Prof. Douglas Harper (Duquesne University, Pittsburg, and IVSA President) and Dr. Alison Rooke (Goldsmiths and CUCR Director). The thesis was supervised by Prof. Caroline Knowles and Prof. David Oswell (both at Goldsmiths).