Since InnoCentive demonstrated the concept, other organizations have arisen to capitalize on outside-in solvers in normally highly specialized fields. Kaggle is like InnoCentive but specifically for posting challenges in the area of machine learning—artificial intelligence designed to teach itself without human intervention.
Shubin Dai, who lives in Changsha, China, was the top-ranked Kaggle solver in the world as of this writing, out of more than forty thousand contributors. His day job is leading a team that processes data for banks, but Kaggle competitions gave him an opportunity to dabble in machine learning. His favorite problems involve human health or nature conservation, like a competition in which he won $30,000 by wielding satellite imagery to distinguish human-caused from natural forest loss in the Amazon. Dai was asked, for a Kaggle blog post, how important domain expertise is for winning competitions. “To be frank, I don’t think we can benefit from domain expertise too much. . . . It’s very hard to win a competition just by using [well-known] methods,” he replied. “We need more creative solutions.”
“The people who win a Kaggle health competition have no medical training, no biology training, and they’re also often not real machine learning experts,” Pedro Domingos, a computer science professor and machine learning researcher, told me. “Knowledge is a double-edged sword. It allows you to do some things, but it also makes you blind to other things that you could do.”
* * *
• • •
Don Swanson saw it coming—the opportunities for people like Bruce Cragin and John Davis, outsiders who merge strands of disparate knowledge. Swanson earned a physics PhD in 1952, and then worked as an industry computer systems analyst, where he became fascinated with organizing information. In 1963, the University of Chicago took a chance on him as dean of the Graduate Library School. As a thirty-eight-year-old from private industry, he was an oddball. The hiring announcement declared, “Swanson is the first physical scientist to head a professional library school in this country.”
Swanson became concerned about increasing specialization, that it would lead to publications that catered only to a very small group of specialists and inhibit creativity. “The disparity between the total quantity of recorded knowledge . . . and the limited human capacity to assimilate it, is not only enormous now but grows unremittingly,” he once said. How can frontiers be pushed, Swanson wondered, if one day it will take a lifetime just to reach them in each specialized domain? In 1960, the U.S. National Library of Medicine used about one hundred unique pairs of terms to index articles. By 2010, it was nearing one hundred thousand. Swanson felt that if this big bang of public knowledge continued apace, subspecialties would be like galaxies, flying away from one another until each is invisible to every other. Given that he knew interdisciplinary problem solving was important, that was a conundrum.
In crisis, Swanson saw opportunity. He realized he could make discoveries by connecting information from scientific articles in subspecialty domains that never cited one another and that had no scientists who worked together. For example, by systematically cross-referencing databases of literature from different disciplines, he uncovered “eleven neglected connections” between magnesium deficiency and migraine research, and proposed that they be tested. All of the information he found was in the public domain; it had just never been connected. “Undiscovered public knowledge,” Swanson called it. In 2012, the American Headache Society and the American Academy of Neurology reviewed all the research on migraine prevention and concluded that magnesium should be considered as a common treatment. The evidence for magnesium was as strong as the evidence for the most common remedies, like ibuprofen.
Swanson wanted to show that areas of specialist literature that never normally overlapped were rife with hidden interdisciplinary treasures waiting to be connected. He created a computer system, Arrowsmith, that helped other users do what he did—devise searches that might turn up distant but relevant sets of scientific articles, and ignited a field of information science that grapples with connecting diverse areas of knowledge, as specialties that can inform one another drift apart.
Swanson passed away in 2012, so I contacted his daughter, political philosophy professor Judy Swanson, to see if she had ever discussed with him his concerns about specialization. When I reached her, she was at a conference, “as it happens, one related to overspecialization in the social sciences,” she told me. From the outside, Judy Swanson looks pretty specialized. Her faculty web page listed forty-four of her articles and books, every single one of which had “Aristotle” in the title. So I asked how she felt about her own specialization, and she seemed surprised. She did not consider herself specialized compared to her peers, she told me, partly because she spends time teaching undergraduates, which requires more than Aristotle. “There is this feeling of frustration,” she told me, “that I should be doing something more specialized.” Academic departments no longer merely fracture naturally into subspecialties, they elevate narrowness as an ideal.
That is counterproductive. As Karim Lakhani put it after his InnoCentive research, a key to creative problem solving is tapping outsiders who use different approaches “so that the ‘home field’ for the problem does not end up constraining the solution.” Sometimes, the home field can be so constrained that a curious outsider is truly the only one who can see the solution.
* * *
• • •
The email subject line caught my eye: “Olympic medalist and muscular dystrophy patient with the same mutation.”
I had just written a book on genetics and athleticism, and figured it would point to some journal article I had missed. Instead, it was a note from the muscular dystrophy patient herself, Jill Viles, a thirty-nine-year-old woman in Iowa. She had an elaborate theory connecting the gene mutation that withered her muscles to those of an Olympic sprinter, and she offered to send more info.
I expected a letter, maybe some news clippings. I got a stack of original family photos, a detailed medical history, and a nineteen-page, bound and illustrated packet that referenced gene mutations by their specific DNA locations. She had done some serious homework.
On page 14 there was a photo of Jill in a blue bikini, blonde hair tousled, smiling and sitting in the sand. Her torso looks normal, but her arms are strikingly skinny, like twigs jabbed into a snowman. Her legs did not look like they could possibly hold her, the thigh no wider than her knee joint.
Beside that photo was one of Priscilla Lopes-Schliep, one of the best sprinters in Canadian history. At the 2008 Olympics in Beijing, she won a bronze medal in the 100-meter hurdles. The juxtaposition was breathtaking. Priscilla is midstride, ropes of muscle winding down her legs, veins bursting from her forearms. She’s like the vision of a superhero a second grader might draw. I could hardly have imagined two women who looked less likely to share a biological blueprint.
In online pictures of Priscilla, Jill recognized something in her own, vastly scrawnier physique—a familiar pattern of missing fat on her limbs. Her theory was that she and Priscilla have the same mutated gene, but because Priscilla doesn’t have muscular dystrophy, her body had found some way “to go around it,” as Jill put it, and was instead making gigantic muscles. If her theory was right, Jill hoped, scientists would want to study her and Priscilla to figure out how to help people with muscles like Jill have muscles a little more toward the Priscilla end of the human physique spectrum. She wanted my help convincing Priscilla to get a genetic test.