The second Computation + Journalism Symposium convened recently at the Georgia Tech College of Computing to ask the broad question: What role does computation have in the practice of journalism today and in the near future? (I was one of its organizers.) The symposium attracted almost 150 participants, both technologists and journalists, to discuss and debate the issues and to forge a multi-disciplinary path forward around that question.
Topics for panels covered the gamut, from precision and data journalism, to verification of visual content, news dissemination on social media, sports and health beats, storytelling with data, longform interfaces, the new economic landscape of content, and the educational needs of aspiring journalists. But what made these sessions and topics really pop was that participants on both sides of the computation and journalism aisle met each other in a conversational format where intersections and differences in the ways they viewed these topics could be teased apart through dialogue. (Videos of the sessions are online.)
While the panelists were all too civilized for any brawls to break out, mixing two disciplines as different as computing and journalism nonetheless did lead to some interesting discussions, divergences, and opportunities that I’d like to explore further here. Keeping these issues top-of-mind should help as this field moves forward.
The following metaphor is not meant to be incendiary, but rather to illuminate two different approaches to tool innovation that seemed apparent at the symposium.
Imagine you live about 10,000 years ago, on the cusp of the Neolithic Revolution. The invention of agriculture is just around the corner. It’s spring and you’re hungry after the long winter. You can start scrounging around for berries and other tasty roots to feed you and your family — or you can stop and try to invent some agricultural implements, tools adapted to your own local crops and soil that could lead to an era of prosperity. If you take the inventive approach, you might fail, and there’s a real chance you’ll starve trying — while foraging will likely guarantee you another year of subsistence life.
It’s certainly fine to use found tools, especially if they solve the problem without a hitch, but too many practicing journalists are still operating in the “forager” mindset. They use only the available tools they can scrounge up to solve the immediate problems related to finding, making sense of, and presenting news information. We heard from David Clinch about how his startup, Storyful, uses readily available tools for vetting and verifying online social media, rather than developing those kinds of technical capabilities in-house. For them, the innovation is in how people use those tools in an overall process. We also heard from Mo Tamman, a veteran data journalist now at Reuters, who initially referred to himself as a “tool whore” but backtracked to settle on the label “tool omnivore,” suggesting he would grab whatever widget would help him nail the story. Nick Lemann, dean of the Columbia University Graduate School of Journalism, called journalists “deadline epistemologists” to communicate the time pressure journalists feel in constructing their knowledge of the world.
Researchers and technologists generally have a different mindset. They engage in forging the implements that let the enterprise scale and flourish. They are often more interested in generalizing and productizing rather than individual one-off stories. Academics just don’t have the same kind of deadline pressure as (most) practicing journalists. And as, I believe, Larry Birnbaum said: Journalists think in terms of stories, technologists think in terms of products. Stories are specific but products are generalizable.
This overarching sentiment seems to lead to the technologists (and researchers) to want to engage with cultivating tools. In the very first session with Irfan Essa and the journalism legend Phil Meyer the message for journalism education was loud and clear: Don’t just teach people how to use tools — teach them how to make them. Imagine a computational fluency and aesthetic akin to the best journalist writing. What might that look like?
This difference in thinking about time horizons opens up a raft of important questions. If journalists should start cultivating computational tools, how should j-schools approach and teach that? Should tools emerge out of universities or newsrooms that subsidize their development, or are startups a better model for innovative tool development? And once you start building tools, how do you make sure the good ones diffuse into widespread practice? Teaching user-centered design is one place to start, as is learning more about computational thinking. And while data and computational literacy skills will increasingly be needed, not all journalists will need to code. The same way there is specialization among other media within the newsroom, building and coding tools for gathering, assessing, and presenting information need not be universal journalistic activities.
Phil Meyer noted that the product of journalism is not eyeballs, but influence. In particular, he noted that trust was the main currency journalists use to achieve that influence. He warned that it’s not enough to just get the information into readers’ hands — it also has to get into their heads. The accountability mission of public affairs reporting is, when you stop to reflect on it, oriented towards influencing the government, in a positive way. It’s all about influence.
Alberto Cairo, when speaking about telling stories with data, brought up a similar idea — how storytelling is often framed as a persuasive, convincing, or impactful enterprise. He suggested that influence might not be a core journalistic value, but it is nonetheless a prevalent one. Indeed, Columbia is beginning to fund research on the study of “impact” — or, more plainly stated, the “influence” — of journalism.
In the session on news dissemination on social media, Gilad Lotan noted the push to quantify influence on social networks. It’s a hard problem, given the myriad exogenous influences that we all experience on a daily basis: friends, social media, television, individual experiences, and so on. There’s money to be made in understanding how or when to present specific content in order to optimize its exposure. And ad revenue from wide exposure is certainly one thing — but other types of influence and impact are less well rewarded by the market, as Duke’s Jay Hamilton alluded to in his session on the media economy.
All of this talk of influence and impact smacks of an opportunity for computational journalism. With growing data sets and better instrumentation across media, could we finally begin building a computational influence engine? Such an engine would know how to optimize not just for revenue, but also for diverse exposure to media and an informed public — as well as for accountability impact. Engineering and biasing computational news media algorithms towards “good” exposure could potentially also help correct for known human biases of perception and cognition.
Several of the speakers at the symposium spoke of the tension between narrative and analytic communication. “We have an impossible and problematic profession because we’re trying to marry narrative and analysis which are fundamentally incompatible…but that’s what makes it fun,” Lemann said. Tamman, the data journalist from Reuters, also spoke of the “analytic spines” that he uses to hold up the narrative part of his data-driven stories. Let’s call this the narrative-dominant frame.
But not everyone wholeheartedly shares the idea that narrative should dominate in the expression of news information. Cairo, a visual journalist who works on data storytelling in infographics, cited a number of objections that people have posed to him over the years: Stories can overweight the value of anecdotes, they can impose narrative on data that is not necessarily complete or cohesive, and they tend to privilege what the designer wants the user to see.
There are other methods that can be used to structure data. Meyer noted that theories are often used to structure the communication of scientific data. And computer scientists are familiar with the idea of using models to describe and communicate data. Models allow for the abstraction of data so that it can be efficiently and effectively computed on. We can call this the analytic-dominant frame.
This brings us back to the raw cultural difference of the value of “theory” or “model” (i.e. understanding the central tendency and abstraction of data) versus the “anecdote” or “outlier” that is so important to journalists feeling they’ve got a good story to tell. We may be just at the beginning of understanding the benefits and tradeoffs of the narrative-dominant frame versus the analytic-dominant frame, but it’s certain that the cultural dilemma of how news communication is approached underscores a central challenge in integrating computation and journalism.
I’ve barely scratched the surface here, and if I’ve answered any questions in this article, it’s probably by accident. If, on the other hand, your curiosity is piqued, then I would urge you to keep your eyes open for the third installment of the Computation + Journalism Symposium in 2014. We hope that, over time, this burgeoning and multi-disciplinary assembly can grow into the place where journalists and technologists come to meet each other and trace a combined path towards a better news information environment. A special thanks goes out to the interdisciplinary organization and sponsorship from Columbia, Duke, Georgia Tech, Northwestern, Stanford, and the National Science Foundation which set the stage with a collective mindshare and ability to convene and attract both computer scientists and journalists.
Painting: “Stone Age,” by Viktor M. Vasnetsov, 1882-85.