The second day of "Compromised Data" starts with Lisa Blackman, who is tracking social media controversies and mapping information contagion. Can we use quantitative methods in non-positivist ways to understand these processes?
Lisa introduces the idea of haunted data, and suggests that we need to think about digital methods as performative: we need to move behind infographics when thinking about visualising data. Part of this is about priming: creating an experimental apparatus that makes people feel that their actions are self-directed, but actually generates such actions through the interventions of the apparatus. Such research is controversial because of its early ties to research into psychic phenomena, however. It is useful, however, to explore information contagion and virality, especially in the context of social media controversies.
Back in the 90s, there was a study which showed that people walked more slowly after being exposed to words related to aging; however, these findings were found to be difficult to replicate, and articles about the failed experiments were not accepted in mainstream journals - eventually, one was published in the open access journal PLOS One. A science journalist further picked up on this story, and his article about it became an intervention in the field. Unsettled issues were revitalised by this study, then.
The journalist's post on his science blog went viral across social media as well as in science magazines, newspapers, and even on Oprah. The information spread for this story is difficult to track and map, therefore; as a researcher you become submerged by the complexity and volume of the data. How might we use commercial software tools to try to follow and map such information contagion?
Bruno Latour has recently addressed software designers in the area of human-computer interaction to develop tools that allow us to remediate more complex forms of analysis. Available tools remediate quite limited forms of content analysis; they don't tell us much, raise issues about sampling and datasets, and are insufficient for modelling the complexity of this controversy. We need tools that allow us to follow and map contagion across multiple media platforms.
The concept of haunted data underpins this project. Data accrues its own life and agencies in this context: the initial PLOS One journal allows people to comment on articles, for example, as a form of post-publication peer review which extends the life of an article (affording it an afterlife) so that data take on a haunted aspect. What we find here is a range of discussions, rumours, gossip, etc., which place the data in new contexts and controversies.
Lisa's intervention in this is to try and develop an affective algorithm that would retweets some of the haunted data that have been edited out of the approved narrative around the controversy. This involves working with automated agents that would retweet some of the data back into the Twittersphere, and this idea is inherited from the Website PopSci. This utilises bots to push back against the diminishment of scientific discourse in recent years.