I developed the concept of “data binds” after I moved to San Francisco, the heart of all things data in the United States, surrounded by friends who work at Google, Facebook, and various tech startups. As much as the HBO show “Silicon Valley” might make us laugh, here it is a reality. One of the billboards that first greeted me on my drive down the traffic-filled Highway 101 was “The Power of Big Data at Work.” Visiting Uber—whose downtown office was full of the unlimited free food and tech optimism you would expect from a startup—an employee explained to me that Uber was tackling some of the most interesting engineering problems, and doing some of the most important work, in the world. And yet, Uber employees were incapable, like so many other Startups, of thinking about the world as anything other than a data or engineering problem.
A few months later, I visited Oakland to look at an apartment for a friend who was about to start a PhD at Berkeley. I drove to the neighborhood, just a few miles from where the new Uber offices would be, and noted the decaying industrial buildings, grass growing through the cracks of the sidewalk, and windows covered in burglar bars. I was struck by how different this was to the shiny, hip image of downtown San Francisco and Silicon Valley, how palpable the income inequality was. Recalling the murals painted on Clarion Alley in the Mission, proclaiming “Evict Google,” how, I asked myself, was the tech industry changing the world here, in Oakland? What kind of “disruption” was going on here?
The concept of data binds was born from a growing concern and fascination with the culture and technologies of big data, and with the politics of knowledge that these produce. In Metabolizing data, I set out to question the effects of data on modern society, and to face my own assumptions and discomfort with digital technologies. That book project drew on my training as a biologist, and on the many years I spent working with post-genomic scientists, to shed light on the opaque inner workings of big data. Through that project, I became fascinated by the data-driven vision of the world being promoted by Silicon Valley: in which the average American spends more than 50 minutes a day on Facebook—more time than exercising and attending social events, and almost as much time as eating.
Data—in all of its diverse forms and effects—is deeply ambivalent. As the world is increasingly oriented around data—in computers, cell phones, apps, fitness trackers—society is faced with increasingly complex questions about whether data is good or bad, or about whether digital innovations enhance or damage our well-being. Data promotes inequality, as tech giants like Amazon deny, until only recently, same-day delivery to mostly black neighborhoods. But data also enables new forms of democracy, as social media plays a role in the Arab Spring and Occupy Movement. How can we make critical sense of data, and its uneven effects on society? Data place us in awkward situations, where there is often not a right or wrong answer. Data have complex effects on people, benefiting some and disadvantaging others, depending on the context. And, ultimately, data require choices that, whether we like it or not, have political implications.
To examine the complex, ambivalent effects of data on the world, we need to attend to “data binds,” to the paradoxical choices and dilemmas people face as they analyze and use data. In data binds, people try to make aspects of the complex, messy, and unpredictable world—be it the human body, the earth’s climate, global inequality—stable and coherent through measurement. Big data gives the illusion of control, but these intractable problems continually elude quantification. This is the data bind: in order to study complex phenomena, researchers must order and simplify them. And as a result, the complex phenomena remain slippery and elusive. (In a fun parallel, “data binding” is also a technical computer science term, and refers to the act of combining two different data elements together, for example between consumers and providers.)
My use of the term data bind draws on the notion of the “double bind,” which was first developed by social scientist Gregory Bateson in the 1950s, and which has been more recently introduced into the social studies of science literature by anthropologist Kim Fortun. In Advocacy After Bhopal, she explores how “enunciatory communities” are called into action in the aftermath of the Bhopal gas disaster through double binds, resulting in social groupings that are never stable, and always negotiating compromises around competing needs. Fortun writes that double binds are “situations that create dual obligations that are related, are of equal value, and yet are incongruent with one another” (Fortun 2001:13). For Fortun, double binds are a way to look at the complexities of social forces, showing how they often cannot be resolves, and how they invoke particular decisions and values.
In Gregory Bateson’s original formulation of the double bind, which he developed through his linguistic work on schizophrenia (see Bateson et al. 1956b; Bateson et al. 1956a; Bateson et al. 1963), the term was a way to think about the complexity of communication during the treatment of mental illness. In his formulation, a double bind might occur when a therapist says to a patient: “I want you to disobey me.” In such cases, to obey is to disobey, and to disobey is to obey, such that no matter what a person does, “he can’t win.” For Bateson, double binds were not only social situations, but also forms of social control, as they were used strategically to exert control with open coercion. In the complex social milieu of schizophrenia, double binds emerged as an important framework for asking how paradoxes were imposed, who imposed them, and how people navigated them.
More than 50 years later, researchers are still encountering double binds—but of an increasingly data-driven and multiple nature. In post-genomic science, data binds happen when researchers develop computational methods to embrace the complexity of biology over space, time, and health, rendered visible through thousands of molecular measurements. But because researchers are constrained by numerous socio-technical problems, they can never make biology fully stable or legible. It is impossible to measure everything, so researchers have to make choices about which aspects of biology to render visible, and which to render invisible. They need, in other words, highly stabilized and formalized computational methods like software and statistics, in order to deal with the multi-causal and dynamic nature of biology.
The same could be said of many other facets of modern life. Global warming, metabolic disorder, flu pandemics: the world’s big problems put the people who seek to study or change them in data binds. This is the tension: in order to study complex phenomena, we have to freeze them. In order to embrace complexity, we have to simplify it. We are continually presented by data binds of varying kinds and intensities, as the concerns of the 21st century continually elude quantification. Faced with large datasets that are impossible to interpret by eye or with intuition, researchers turn to data-driven techniques to explore the complexity of their data. Big data, in this way, is a self-fulfilling prophecy, as it drives the collection of more and more data.
Ultimately, data binds are a tool for examining the challenges and strategic ways in which complexity is frozen and rendered quantifiable. Examining the strategic ways in which researchers reify complexity, simplifying it, shows how certain aspects of our social world are made legible, and by what means. It shows how the 21st century practices of quantification entail choices—some overt, some not—about which aspects of life to make visible, and which to silence. Ultimately, data binds are a means to interrogate the politics that shape how science gets done. Examining the moments in which the researchers finds themselves in data binds—and the ways in which they articulate and navigate them—reflects the contexts in which data binds arise, the institutions and forces that enable and reinforce big data. Thinking about how data binds are imposed, and what people and institutions impose them, reveals the values and norms of data-intensive science.
The above text is an adapted excerpt from my book Metabolizing Data, where I first developed the idea of the data bind.