Yesterday a Facebook friend of mine wrote about an “eerie and creepy” experience. An advertisement had just appeared in the margin of the Facebook page she was reading, showing a tee-shirt with the lettering, “Just a California Girl in a New Jersey World.” It is no coincidence that my friend is a native of California and now lives in Flemington, New Jersey. Facebook obviously sold this user-profile information to advertisers, one of whom found a way to develop a highly personalized product.
But my friend did not buy the tee-shirt, and if enough other targets of this pitch find their personalized tee-shirt unappealing, the manufacturer will lose money on the ad campaign (plus the costs of tooling up to produce the tee-shirts; the manufacturer surely doesn’t have an inventory of shirts for every possible two-state match-up). What this tells me is that although Facebook’s data-gathering (a robotic function) makes it possible to create highly personalized products, the actual creation of products that people will buy remains a human function. In the brave new world of Big Data, human creativity is still needed to make a sale.
Something very similar became clear in this month’s congressional elections. As The New York Times pointed out,
Modern political campaigns home in on their key voters with drone-like precision, down to the smallest niche — like Prius-driving single women in Northern Virginia who care about energy issues. They compile hundreds of pieces of data on individuals, from party registration to pet ownership to favorite TV shows. And they can reach people through Facebook, Pandora, Twitter, YouTube or cable television.
The only problem: They do not have enough messages for them all.
The Big Data era of politics has left some campaigns drowning in their own sophisticated advances. They simply cannot produce enough new, effective messages to keep up with the surgical targeting that the data and analytics now allow.
…Or, as Joe Rospars, the founder of the Democratic digital agency and technology firm Blue State Digital, put it, “The science is ahead of the art.” An analytics team can help a campaign make “a much more targeted buy,” he explained, but that alone will not offer a particularly efficient return on investment if the ad is still “just a white guy in a suit.”
As a result, the people who design the advertisements for electoral campaigns end up trying to tease out certain large slices of voters with something in common, such as “soccer moms” or “angry white males.” The campaigns do not have enough creative people to craft the highly personalized ads that should be possible given data-analysis tools.
On a recent broadcast of NPR’s “On the Media” (sorry, I can’t remember which date), I found another example of the limitations of technology. You may remember that YouTube moved quickly to take down the videos of the recent beheadings in Syria. You may not know that YouTube takes down many other videos because of pornographic or sadistic content, and so do most photo-posting sites, such as Flickr and Photobucket. Computer technology makes it easy to post photos and videos, and to make them searchable by keywords, but the explosion of content that has appeared on the Web requires human eyes to decide which photos and videos violate site policies.
One indication of how subtle the human decision-making must be is the fact that photo sites prefer to offshore this work to the Philippines rather than to India, where the workers are lower-paid. The reason is that Filipino workers have a better understanding of American culture and therefore can decide whether (for example) a shot of someone in a bikini is too revealing, whereas an Indian worker might reject every bikini shot.
The takeaway is that technology sometimes creates a need for more human workers, and not just those employed in creating, manufacturing, or repairing the technology. It sometimes creates opportunities for work that requires great creativity or subtle judgments. Robots may be driving cars, but they are not yet metaphorically in the driver’s seat.