Sensors and Journalism

The Landscape for Sensors and Journalism

Throughout 2013 and 2014, whenever the Tow Center gathered people together to work on the topic of sensor journalism, we needed to set the scene. When it comes to labels, “sensor journalism” isn’t even as understood as “data journalism.” So, although the last thing journalism needs is a new term to define a fragment of its practice, we must draw some boundaries and describe the landscape to bring readers of this report into a common language. Nonetheless, we reserve the right to retreat or advance from this ground, or even open the borders, as seems necessary in the rapidly evolving world of journalism circa 2014. By erecting a frame of reference for journalistic sensing we hope to help readers efficiently understand which intellectual tools they can apply to their own sensing and sensor-based work by others. However, by describing this field discretely we do not mean to advocate that it be practiced discretely from other types of reporting; quite the opposite. Indeed, the legal and ethical sections and most of the case studies that follow demonstrate the value in combining sensor-based reporting with other journalistic tools, including personal interviews and shoe-leather reporting, so that the data can be incorporated with context, narrative and emotion.

King Data and the Wide Angle for Journalism First, we offer some observations about why it might be worth focusing on sensors in combination with journalism. This is the context for this report; a suggestion of why there is value in this particular vein of research. Sensors are a way of collecting information about the world. Journalists trade in acquiring information, analyzing it, organizing it, and distributing it. That alone suggests a natural fit. However, beyond that, journalists are currently paying special attention to data that can be easily parsed by a computer. Many readers of this report will be familiar with the impact of data-specialist teams put together by the top tier of American and international news companies, whether in longestablished newsrooms like The New York Times, the LA Times, the Guardian, and the Washington Post, or ambitious newcomers like FiveThirtyEight, Buzzfeed, and Vox.1schools have recruited professors who teach data analytics, principles, and presentations. Hundreds of Master of Science students entering Columbia University’s Graduate School of Journalism in 2013 took at least three seminars about data practices. Many go on to learn Python and R programming languages in greater depth. The primary U.S. conference for data journalism, NICAR, went from a conference of a few hundred people in 2010 to one with a thousand attendees in 2014. The rise of data journalism has coincided with an age in which technology is becoming cheaper, more capable, and more widespread. Many observers would suggest a causal relationship: When computers permeated more homes and schools, more children learned programming skills. More systems and data about the world have been digitized. There are more stories to be found in databases and more journalists working in the profession with the interest and skills to find them. Sensors Everywhere The hardware components of technology, including sensors, are also cheaper and more ubiquitous. While the following examples of sensors don’t encompass the breadth of technology observed in this report, they can be a useful illustration. Cellphones include cameras, accelerometers, and GPS sensors, microphones and radio frequency receivers. Those components now cost a few dollars at wholesale. Young children (or at least their parents) can spend $36 to buy a pair of sneakers that sense movement and trigger small lights in the soles. Sensors are baked into civic processes like traffic control, and industrial processes like stock control. A private company called Planet Labs has put 28 toaster-sized satellites into orbit, designed to operate as a flock of cameras pointed toward the earth. It’s aiming to have an array of 100 craft in space by March of 2015. We are living in a sensed world. Aside from the sensors incorporated into finished products for consumers, governments, and private enterprise, sensors comprise a key component class used by the “maker movement.” That ecosystem encompasses electronics prototyping platforms such as Arduino and Raspberry Pie, DIY electronics retailers Sparkfun and Adafruit, and, if interpreted widely, also includes KickStarter’s crowd-sourced product development and the consumer-side of 3D printing. Taken together with the spread in programming skills, it is fair to say that DIY hardware development is flourishing. In early 2013 the makers of the popular electronic prototyping platform, Arduino, said there were 700,000 boards in operation. They estimate that the number doubles if one includes the clones (which are legal, given their open-source license).

So for journalism, there is a special symmetry of demand and supply. Behind the computer-aided reporters and data savvy journalists have come a generation of programmer-journalists. They all compete for fresh data to include in their stories. Another artifact of the digital-first era is the development of news apps and interactive news graphics, in which users’ ability to explore data requires that information be available in granular form, not just as a summary paragraph or a static graphic (although those are also both perfectly justifiable outcomes of a data-reporting process). As the case studies in this report explore, sensors can produce the data demanded by computer-aided journalistic processes. However, beyond simply satisfying the existing demand for data, particular characteristics of modern sensors and their accompanying technologies produce opportunities for new reporting processes. The low cost of some sensor types enables experimentation and new modes of use. In the case studies that follow, we see examples of cameras being practically disposable, or at least sacrificial. Environmental sensors might be widely deployed and left always-on. Cameras can be used, not for the scenes they depict, but as a source for a pixel-by-pixel computational analysis. The sound waves picked up by a microphone have been parsed for the characteristic signature of a newsworthy event. Why Would Journalists Want to Sense? When the Tow Center ran a workshop (in June of 2013), hosting a range of journalists, researchers, and technologists we asked participants about their hopes and ideas for sensors in journalism. A couple of themes emerged: The first was simply a desire for more data to use in their reporting. Especially in the environmental sphere, our participants felt there was a deficit in the data being provided by official sources. They wanted data with better spatial coverage—often targeted where they expected to find problems but couldn’t be sure. Pollution monitoring near industrial facilities was the most common example. That desirability of having data about more places can logi cally be extended to increased temporal resolution. Instead of taking a sample once a month or once a week, some journalists want to monitor aspects of the world all the time.2examples of sensors used to collect with greater spatial and temporal density—not just on environmental topics. However, as well as getting data with more resolution, the case studies show imaginative applications of sensors providing different types of information, especially around the location of people over time. We heard another potential benefit of using sensors in journalism: to take human observations and impressions and make them specific, so that they might be used for comparisons. Often that meant quantifying an observation: The amount of a chemical in the air matters if it is to be compared to a known health risk factor; the speed of a car matters if it is to be compared to a law. But journalists also want to make comparisons across time and space; does one neighborhood in Washington have more gunshots than the next? Has the number gone up or down over the last year? A sensor can record an aspect of the world so that it can be specified and transparently communicated. These aspects, we believe, are the context for our research into sensors and journalism. We are in an era in which reporters are hungry for data, and increasingly expert in using it; in an age when sensing technology is developing radically and permeating every aspect of modern life. Those trends, taken together in journalism, have produced new demands that sensors might meet, opportunities that might be exploited, and benefits that might be realized.

Journalistic Sensing in This Report The case studies in this report document journalistic projects using sensors that clearly fit within these trends. An early classic in participatory electronic sensing was The Cicada Tracker, which saw WNYC listeners build Arduino-based sensors to contribute readings of temperature readings in their local environment. USA Today’s multi-year effort to take almost a thousand soil samples became the Ghost Factories project, and picked up numerous investigative journalism awards. Nonetheless, through our discussions and work over the last year, we’ve found ourselves pausing on specific journalistic projects and asking whether or not they fit into the ’sensor journalism’ field. Again, we have no desire to make delineations for their own sake, but simply to demonstrate how this framework applies. We’ve worked through a few examples to help readers understand why they are reading various chapters that follow. Drones seem to be worth including in this report. To date, most journalistic uses of drones have been for collecting photos and videos (although other industries also leverage drones’ ability to collect 3D landscape data and environmental data). While cameras have been used in journalism since they were invented, the qualifying aspect for drones, we believe, is that they leverage the radical advances in camera miniaturization. Insofar as drones carry sensors and thereby extend the reach of reporters to observe and record the world, they fit into this field. Also, through the second half of 2013 and the start of 2014, civic uses of drones—including journalism— have become the topic of increasing research, mainstream media attention and regulatory action. Many newsrooms we spoke to have started planning for drone use. We think including drones in this report makes it more relevant to an urgent conversation in the news industry.

Likewise, we have included journalists’ use of data produced by sensors they didn’t commission or control. There’s a good counterargument to be made that this isn’t distinguishable from data journalism, but this report is not trying to carve out a field of sensor journalism that is apart from data journalism. We refer again to our goals for this report: to help journalists use these sensors as well as they can, and to help them understand data that comes from sensors. While some journalistic uses of sensors do involve building customized sensors, or running their own sensing programs, we see no reason to narrow the discussion to those use cases. Two of the case studies deal with data that was released when journalists asked for data under Freedom of Information Act principles. Although we haven’t included a case study about journalists’ use of remote sensors on satellites, we know of one newsroom investing a lot of time to understand and use that data for its work. In the coming months, readers can expect to see more stories based on innovative uses of remote sensing. Data Collection Beyond our Purview On the other hand, there are some related journalistic practices we’ve left out, even though they may share characteristics with the history, theories, and journalistic uses of sensing. In the process of drafting his paper on the epistemological considerations of sensing, University of Wisconsin assistant professor Lucas Graves wrote a provocation to include polling. Like sensing, polls are a way for journalists to systematically collect information about the world, often in a form ready for computation. Bloomberg and Reuters use consumer and business sentiment polling on a weekly basis. They brand strategically important polls and form partnerships with polling organizations and universities. Likewise, during election seasons (are there any others?), polls commissioned by the Washington Post and ABC News move to the top of the news agenda, along with those performed by Rasmussen and Gallup. So, there’s a strong argument for ushering polls into our work. However, the origin of polling data

is a human interaction, which seems to make the practices of polling and sensing distinct. That said, we do acknowledge human agency in designing sensors, sensor data collection methods, and the analytic processes working with sensor data. Likewise, humans as sensors have been excluded from this report. Although we accept absolutely that humans make observations about the world through any combination of their five senses and can record the information, the lack of a mechanical process that can be interrogated and reproduced would seem to separate human sensing from technical sensing. Once again, we acknowledge counterarguments; advances in social science experimental techniques have made human observations more reproducible, while technical and mechanical sensors inherit design decisions influenced by human subjectivity. Nonetheless, human observations seem to have a different degree of controllability and specificity and are not as influenced by the macro-trends outlined above. For those reasons, we’re not researching human sensing. Perhaps the hardest exclusion has been software sensing. When marketing firms examine bit-torrent networks to research the popularity of movies and albums, their practice shares many characteristics with physical sensing. It is another intersection of new technologies with the demand to make observations about the world. It can produce massive amounts of interrogatable data. It can produce real-time information or information to be stored, processed, and disseminated. Still, there are differences as well; software sensing seems to be more concerned with the virtual world, whereas the practices we’re interested in here are more about observations of the physical world. But again that might be a false distinction: In our case studies we have examples of physical observations moving immediately into digital, networked information.

So, our borders of convenience that exclude these journalistic practices could easily be redrawn to welcome them in. Some of the characteristics of sensors we describe are shared with these practices we’ve left out. Some of the legal and ethical considerations of reporting with sensors apply, and the lessons and observations found in the case studies may be just as relevant. So, if you find any of our observations and lessons about sensing useful for other practices, take them with our blessing.