Of Algorithms and Audiences
After attending two very different conferences over the course of a week to talk about the same digital research project, I found myself in the old awkward position of “desperately seeking the audience”—of computational tools and digital methods for media studies research.
At the end of October, I traveled to the 2014 IEEE (Institute of Electrical and Electronics Engineers) International Conference on Big Data in Washington, DC (slides and paper available here). The second conference, Film and History, was a bit closer to home, both literally—the conference hotel was five miles from my house—and in terms of the disciplinary concerns of researchers. At both, I was presenting material based on research developed from work on Project Arclight, the winner of National Endowment for the Humanities’s Digging into Data Round Three grant funded by the Institute of Museum and Library Services and Social Sciences and Humanities Research Council. Much of our current work with Arclight focuses on the creation of an online application that will enable researchers to track terms and trends throughout a corpus of 2 million pages of digitized film and broadcasting trade journals, magazines, and books. Yet we hope that the project will serve as a broader catalyst for building connections between media studies and digital humanities efforts.
At the first conference, between sessions and over meals, I spoke with several researchers struggling with issues of tool adoption. After a full day of presentations describing innovative and powerful new tools built from the collaborations of dozens of scholars across disciplines, the question remained: how to generate excitement about these projects that could incite scholars and students to use them? And, for those presenters coming from a computer sciences background, is this really what discipline-area scholars want from digital tools?
Film and History was instructive. At a special event workshop on historical methods shared with the Wisconsin Center for Film and Theater Research, my fellow Arclight team members and I presented a brief introduction to digital analytics. It became clear that—at least for those in the room—the barriers to adoption weren’t a lack of interest, but a lack of familiarity with some of the major precepts and possibilities of digital work. For those suspicious of humanities inquiry bound to binaristic or quantitative frames, we were able to describe the many ongoing conversations in digital humanities regarding measures of uncertainty, the inevitability and impact of human intervention in computational processes, and the consequences of adapting tools designed for other scientific and business purposes for humanities work. For those unfamiliar with the field’s current key methods, such as topic modeling, we used software demonstrations to showcase their capabilities and limitations. As a follow-up to attendee questions, Eric Hoyt created a tutorial on topic modeling for media history using the Media History Digital Library corpus, which he posted on the Arclight website. The workshop also enabled us to hear what scholars hoped digital tools might accomplish and guided our attention to capabilities we might incorporate in our own tool development work. While the workshop model seems to function well on a small scale, allowing us to respond to the individual concerns of those already somewhat interested in computational methods, how we might broaden the appeal of such workshops remains to be seen.
There are a number of reasons why this expansion is urgent, a few of which I’ll mention here. First, and most importantly, conversations in digital humanities have been invaluable for demonstrating the extent to which all of our work—whether we consider ourselves to be using computational methods or not—is constituted by digital technologies. Representations recently published an excellent forum on full text search—what we might consider a rather banal, quotidian tool—and the consequences of not understanding the politics of search algorithms. (Ted Underwood’s “Theorizing Research Practices we Forgot to Theorize Twenty Years Ago” is particularly insightful in its discussion of how digital technologies are structural to academic research and have been for some time.) Second, as media scholars begin to use computational methods to serve their existing research agendas, the peer review process will be in need of people who can critically assess the quality and contributions of technological methods. Third, computational analytics, digital collaboration strategies, and the online distribution of scholarly work could provide useful additions to graduate methods courses, enabling future scholars to put these methods in conversation with existing scholarly practices in new and useful ways.
For our part, we’re hoping that the Arclight website will become a useful resource for those interested in the pairing of digital methods and media studies, but we’d also like to find other avenues to make our work appealing and accessible. While this might take the form of more conference workshops, Skype seminars, and classroom visits, we’d be interested in hearing any suggestions or questions you might have in the comments below.