In my last post, I discussed the close relationship between the business model of BuzzFeed and online production and performances of identity. In this post, I’ll explore the system of human and algorithmic processes that power the cycle of “12 Reasons Early ’00s Christina Aguilera Was Before Her Time” and “These People Tried Weird Milk For The First Time” that we see dance across our Facebook feeds day in and day out.
First, a little history: BuzzFeed didn’t start as a publishing outlet, but as an experiment in tracking and aggregating viral content. As a New York magazine article recalls, “originally, BuzzFeed employed no writers or editors, just an algorithm to cull stories from around the web that were showing stirrings of virality.” As BuzzFeed evolved, its leadership realized that it could use what it was learning to create original content, and it became a “lab” for experimenting with creating highly shareable content. These experiments have drawn from the idea, outlined by BuzzFeed founder Jonah Peretti and Duncan Watts in a 2007 Harvard Business Review article, of “big-seed marketing, which combines viral-marketing tools with old-fashioned mass media in a way that yields far more predictable results than ‘purely’ viral approaches like word-of-mouth marketing.” Led by Peretti (who New York calls a “semiotic Darwinist”), BuzzFeed has continued to expand its original content and has positioned itself as a platform into which shareability expertise has been built, ready to help messages – whether that message is a quiz on how well you know a TV theme song or a paid message from your friends at Taco Bell – travel throughout the web.
As I discussed in my last post, BuzzFeed’s current iteration in many ways seems like an “optimization” of this model for the social web, as its staff produces the kinds of content people are likely to share with their online networks. BuzzFeed further ups the “rate of reproduction” of its content by providing tools that make it easy to engage and share – like the prompt to register your reaction as “cute” or “fail” or the buttons positioned on each page, linked to various social networks. All of this both promotes mobility throughout the social web and provides a whole lot of data. The possibilities afforded by this system can be seen especially clearly in BuzzFeed data scientist Ky Harlin’s discussion of machine learning as applied to BuzzFeed’s most (in)famous format, the listicle: “A list is in itself a mini pool of content….So we’ll try to really figure out what people are engaging with and turn a list of 45 items to a list of 25 items without the duds, reordered to make it most likely to share.” A listicle, in this understanding, is broken into discrete units and data points.
In many ways, this exemplifies the modes of processing information that underlie digital systems. In Keywords for American Cultural Studies, Tara McPherson discusses the distinction between analog and digital understandings of information, writing:
Digital computers….process digital data as discrete units called bits, the zeroes and ones of binary code. A transistor in a digital computer has two states, on or off; a capacitor in an analog computer represents a continuous variable. The digital privileges the discrete and the modular; the analog represents continuity. As humans, we perceive the world analogically, as a series of continuous gradations of color, sound, and tastes.
Here we can see the intersection at which BuzzFeed operates: its bread and butter is not just human identity – which would seem to imply the continuity associated with analog human experience – but human identity as produced and performed online. In spaces built, maintained, and tracked using digital logic, through which pathways are partially directed and monitored by algorithms, the human and machine (and the analog and the digital) become increasingly difficult to separate. In “The Algorithms of Our Lives,” Lev Manovich writes, “Software has become a universal language, the interface to our imagination and the world.” This rings true in relation to the algorithms that track our activity on BuzzFeed or deliver us updates on Facebook, but in the hype around “digital everything,” it can be easy to forget that these algorithms exist as part of a larger network. BuzzFeed content is “engaged with” not just through the button clicks its algorithms can register, but also in quiet smiles or water cooler conversation, and these various modes are together woven into a highly interconnected network of analog/digital experience.
Just as algorithmic analysis of BuzzFeed’s content exists as part of a complex system that involves both humans and machines, so too do the processes underlying the algorithm’s creation and deployment. In a recent piece for The Atlantic, Ian Bogost criticizes the idea of the all-powerful algorithm, placing our use of algorithms as metaphor for contemporary life in a history of “explaining the present via the output of industry” and arguing that “the algorithmic metaphor gives us a distorted, theological view of computational action” that skims over the complex mix of parts, from servers to security guards, that underlie any algorithm. He points to examples like Netflix, which has recommendations that may seem cleanly “algorithmically generated” but in fact reflect the human labor of watching films and tagging them with metadata.
Discussion of BuzzFeed often centers on its algorithm, using much of the language of automation that Bogost critiques, but BuzzFeed’s current success reflects a system – built in response to the question that Peretti phrases as, “If you actually have an idea or an insight into how things are shared or why they’re shared or what works, can you build that into either the technology platform or the data science or the culture of a team of people?” – of which the algorithm is just one element. BuzzFeed is growing its human labor force, with human workers both responding to outputs and generating inputs in tandem with algorithmic labor. In a 2014 Q&A, Peretti discussed BuzzFeed’s news story generation process, saying, “None of that is directly tied to any metrics. Although having seen what people’s response is in the past, it actually might give people a gut feeling.” This idea – a "gut feeling" drawn both from personal, analog data on stories that will resonate and from digital data on clicks and shares and virality – might just sum up BuzzFeed’s positioning as a platform built and sustained on a network that spans analog and digital understandings: the “distributed cognition” of networked life.