Dr. Dan Angus (School of Communication and Arts, UQ)
Assoc. Prof. Adrian Athique (Institute for Advanced Studies in the Humanities, UQ)
Assoc. Prof. Craig Hight (School of Creative Arts, UON)
Dr. Brian Yecies (School of Humanities and Social Inquiry, UOW)
The Dynamics and Potentials of Big Data in Audience Research
This paper considers the dynamics of the big data paradigm and its potentials for scholars working in audience research. Media dynamics are determined in this instance as the presumptions, imperatives and motives that shape the paradigm itself, along with the interaction of institutional forces at play in the utilisation of audience data. With the ground set by those dynamics, the media potentials of big data procedures emerge from both the applications and the possible outcomes of these techniques. This paper makes the assertion that the two primary motivations of the big data paradigm stem from two longstanding preoccupations of human science, namely numerology and alchemy. Upon the basis of this critique, I will seek to consider the broad implications of the increasing centrality of big data in audience research.
Designing media algorithms: Insights from the creation of a content recommendation system
Ten years ago, Netflix launched a US$1mil prize for any person or team that could beat its own Cinematch content recommendation algorithm by a large margin of improvement. Technically known as collaborative filtering, these algorithms use known user ratings data to train their recommendation making mechanisms, so called ‘training data’. If the recommendations made by the trained algorithm on previously unseen ‘test data’ are good, i.e., if users rate previously unseen films recommended to them highly, then the algorithm is scored favourably. The assumption behind such systems is that a users’ tastes are predictable and will closely match those of similar peers in the system, however lacking in subtlety this may seem. The Netflix prize is a prime example of the classical computer science ‘train/test’ paradigm, where a quantified gold standard exists for engineers to try to match or beat. In this talk I will outline this dominant evaluative paradigm, and offer a critical evaluation of the ways it is impacting our cultural sphere using examples from social media, internet search and music and film recommendation.
Audience research in an era of Big Data: from the outside, looking in
Manovich has used the notion of ‘data classes’ to suggest the inequalities of access to datastreams which are inherent to the practices of Big Data. While a small elite have access to vast streams of data generated from a rapidly expanding sources, the overwhelming majority of digital users are (often unwitting) data generators, leaving digital traces through most of our online activities. At a time when the largest social media platforms have proprietary access to data collated from the majority of online populations, researchers are typically faced with the possibility of an ocean of data on audience practices just beyond their reach. Unless researchers are on the ‘inside’ of companies and organisations harvesting such material, part of a small elite of privileged collaborators, they are on the ‘outside’ forced to engage with the limitations and constraints afforded by these platforms at a variety of levels.
This presentation addresses some implications for audience research suggested by a case study in data scraping the YouTube platform. YouTube presents challenges derived from a dependence on the information generated by users themselves, the manner in which this is structured and surfaced by the platform’s automated mechanisms, the broader drift of YouTube toward a commercialist ethos, and the data policies coded in its application program interfaces (APIs). Audience researchers working on such platforms need to not only continually refine computer-based techniques for accessing and collating material, but, perhaps more crucially, to develop a critical understanding of the epistemological implications of audience research practices operating in this environment.
Chinese Transnational Media Audiences and ‘Good’ Data Mining on Douban OSN
This talk presents preliminary findings on an investigation of user-generated content from a popular online social network in China: Douban, where media-savvy ‘digital natives’ post and follow comments about films, tv shows, books, music, and current events. Topical conversations include a plethora of comments on a wide range of domestic, international and transnational films that are encouraging audiences to form new views about genre, film production, and global cultures more generally. Included in the mix are a number of South Korean films that have become a welcomed alternative to Hollywood blockbusters, particularly given their star power, fresh aesthetics and creative stories for which Korean cinema has become known since the Korean government relaxed film censorship in 1996. How are Chinese audiences on Douban OSN making sense of this increasing influx of popular culture in China? How can we apply so-called Big Data methods to transdisciplinary studies in humanities? To gain deeper insights into this emerging arena, this talk addresses some of the challenges and benefits of harvesting ‘good’ rather than simply ‘big’ data samples, and offers some lessons about managing data involving films, users, comments and their interconnected relationships in Chinese online social networks.
Friday May 27th, 2016 9.00am
Seminar Room, Level 4 Forgan Smith Tower, University of Queensland, St Lucia
R.S.V.P. to Jill Paxton @ email@example.com advising any dietary requirements as Morning Tea will be provided