Whereas Chapter 6 explored the theory of collective intelligence, Chapter 7 looks at how collective intelligence is being employed in real-life crowdsourcing examples.
In the late fall of 2004 Karim Lakhani, a PhD candidate at MIT’s Sloan School of Management, was suffering from a common affliction among graduate students. “I hit that point where I just couldn’t stand to spend any more time on my own dissertation,” recalls Lakhani. He’d been analyzing how innovations emerge in open source software, but after four years of work, he was burnt out. It was time for an extended vacation. “I stopped everything and read Neal Stephenson’s Baroque Cycle.” Stephenson’s trilogy is a work of historical fiction about the Age of Enlightenment in Europe, and it made a big impact on Lakhani. “It’s all about the establishment of the Royal Society and the dawn of rational thinking about the world and the invention of calculus.” By Lakhani’s lights, the Baroque Cycle was a narrative history of innovation.
One passage in particular caught Lakhani’s attention. Stephenson related the true story of the longitude prize. In 1714 the British established a commission offering 20,000 pounds (roughly $6 million today) to anyone who could solve the mystery of longitude. The Royal Navy’s inability to measure lateral progress across the oceans had resulted in the loss of untold numbers of ships and their cargoes, causing a serious drain on government finances. “The top scientific minds of that time, including Isaac Newton, had tried to find a device that could determine longitude, and none succeeded,” notes Lakhani. The solution—a clock that operated with superb accuracy, even during the rigors of an overseas voyage—was developed by John Harrison, an uneducated cabinetmaker from Yorkshire. “I read that and thought, ‘Huh. Kind of like open source. Someone poses a problem and all sorts of random, strange people show up and say I’ve got an answer for you, and it’s never the one you’d anticipate.” The longitude prize constitutes the earliest known example of crowdcasting—broadcasting a problem to the widest possible audience in the blind hope that someone, somewhere—maybe even a Yorkshire cabinetmaker—will come up with a solution.
Lakhani returned to his dissertation, but now he was determined to look at innovation with a much broader view. He’d heard of the problem-solving scientists at InnoCentive, and wondered whether the company wasn’t a modern-day version of the Longitude Prize. “I told them I wanted to figure out how their problems were solved. They were thrilled.” Lakhani teamed up with a PhD candidate at the Copenhagen Business School and two InnoCentive scientists. Over the course of the next year, they looked at 166 scientific problems afflicting the R&D labs at 26 separate firms. In the summer of 2007 they published their findings as a Harvard Business School working paper . The results ran counter to decades of conventional wisdom in the sciences, but they probably wouldn’t have surprised John Harrison.
A Whole New Paradigm
The future of corporate R&D can be found above Kelly’s Auto Body on Shanty Bay Road in Barrie, Ontario. This is where Ed Melcarek, 59, keeps his “weekend crash pad,” a one-bedroom apartment littered with amplifiers, a guitar, electrical transducers, two desktop computers, a trumpet, half of a pontoon boat, and enough electric gizmos to stock a RadioShack. On most Saturdays, Melcarek comes in, pours himself a St. Remy, lights a Player cigarette, and attacks problems that have stumped some of the best corporate scientists at Fortune 500 companies.
Not everyone in the crowd wants to make T-shirts. Some have the kind of scientific talent and expertise that used to exist only in rarefied academic environments. In the process, forward-thinking companies are changing the face of R&D. Exit the white lab coats; enter Melcarek, who like the chemist Giorgia Sgargetta, is one of the 140,000 “solvers” who make up the network of scientists on InnoCentive. Previously I looked at InnoCentive’s ability to harness the excess capacity of people like Melcarek and Sgargetta. Now it’s time to take a deeper look at InnoCentive’s methodology, as it illustrates the ideas laid out in the previous chapter.
Pharmaceutical maker Eli Lilly funded InnoCentive’s launch in 2001 as a way to connect with brainpower outside the company – people who could help develop drugs and speed them to market. From the outset, InnoCentive threw open the doors to other firms eager to access the network’s trove of ad hoc experts. Companies like Boeing, DuPont, and of course, Procter & Gamble now post their most ornery scientific problems on InnoCentive’s Website; anyone on InnoCentive’s network can take a shot at cracking them. The companies generally pay solvers anywhere from $10,000 to $100,000 per solution. (They also pay InnoCentive a fee to participate.) Jill Panetta, InnoCentive’s chief scientific officer, says more than 30 percent of the problems posted on the site have been cracked, “which is 30 percent more than would have been solved using a traditional, in-house approach.”
“Everyone I talk to is facing a similar issue in regards to R&D,” says Larry Huston, Procter & Gamble’s former vice president of innovation and knowledge. “Every year research budgets increase at a faster rate than sales. The current R&D model is broken.” P&G is one of InnoCentive’s earliest and best customers, but the company works with other crowdsourcing networks as well. YourEncore, for example, allows companies to find and hire retired scientists for one-off assignments. NineSigma is an online marketplace for innovations, matching seeker companies with solvers in a marketplace similar to InnoCentive. “People mistake this for outsourcing, which it most definitely is not,” Huston says. “Outsourcing is when I hire someone to perform a service and they do it and that’s the end of the relationship. That’s not much different from the way employment has worked throughout the ages. We’re talking about bringing people in from outside and involving them in this broadly creative, collaborative process.”
InnoCentive’s solvers are not who you’d expect. Many are hobbyists working from their proverbial garage, like the University of Dallas undergrad who came up with a chemical to use in art restoration, or the North Carolina, patent lawyer who devised a novel way to mix large batches of chemical compounds. Or in Melcarek’s case, a mildly eccentric electrical engineer whose lab doubles as a music studio. Yet Melcarek solved a problem that had stumped the in-house researchers at Colgate-Palmolive. The giant packaged goods company needed a way to inject fluoride powder into a toothpaste tube without it dispersing into the surrounding air. Melcarek knew he had a solution by the time he’d finished reading the challenge: Impart an electric charge to the powder while grounding the tube. The positively charged fluoride particles would be attracted to the tube without any significant dispersion.
“It was really a very simple solution,” says Melcarek. Why hadn’t Colgate thought of it? “They’re probably test tube guys without any training in physics.” Melcarek earned $25,000 for his efforts. Paying Colgate-Palmolive’s R&D staff to produce the same solution could have cost several times that amount – if they even solved it at all. Melcarek says he was elated to win. “These are rocket-science challenges,” he says. “It really reinforced my confidence in what I can do.”
Melcarek, who favors thick sweaters and a floppy fishing hat, has charted an unconventional course through the sciences. He spent four years earning his master’s degree at the world-class particle accelerator in Vancouver, British Columbia, but decided against pursuing a PhD. “I had an offer from the private sector,” he says, then pauses. “I really needed the money.” A succession of “unsatisfying” engineering jobs followed, none of which fully exploited Melcarek’s scientific training or his need to tinker. “I’m not at my best in a 9-to-5 environment,” he says. Working sporadically, he has designed products like heating vents and industrial spray-painting robots. Not every quick and curious intellect can land a plum research post at a university or privately funded lab. Some must make HVAC systems.
InnoCentive has been a ticket out of this scientific backwater for Melcarek. For the past three years, he has logged onto the network’s Web site a few times a week to look at new challenges. Until recently these had been categorized as either chemistry or biology problems. Melcarek has formal training in neither discipline, but he quickly realized this didn’t hinder him. “I saw that a lot of the chemistry challenges could be solved using electromechanical processes I was familiar with from particle physics,” he says. Besides the fluoride injection challenge, Melcarek also devised a successful method for purifying silicone-based solvents. That challenge paid $10,000. Melcarek has since gone on to win five additional InnoCentive challenges. “Not bad for a few weeks’ work,” he says with a chuckle.
Melcarek has discovered something of a winning formula: Find chemistry or biology problems that he can crack using his background in physics and electrical engineering. In 2007 InnoCentive launched a category for engineering challenges, but Melcarek doesn’t bother with it. All seven of the problems Melcarek has solved were in other fields.
This says a little something about Melcarek (he’s a man who likes to work off hunches), but it says a lot more about InnoCentive. When Lakhani dug into InnoCentive’s data, he discovered that Melcarek wasn’t the exception, he was the rule—the scientists most likely to solve a problem were the ones you’d least expect to be capable of solving it. “We actually found the odds of a solver’s success increased in fields in which they had no formal expertise,” Lakhani says. The further the problem was from their specialized knowledge, the more likely it was to be solved. “Think of the problem as a flower. Except the goal is to attract not only the most insects, but the most diverse group of insects.”
And Lakhani’s paper contained an even more interesting gem: A full 75 percent of successful solvers already knew the solution to the problem. The solutions to the problems in the study—many of which, recall, had stumped the best corporate scientists in the world after years of effort—didn’t require a breakthrough, or additional brainpower, or a more talented scientist’s attention; they just needed a diverse enough set of minds to have a go at them. It would seem to be evidence that Hayek was right: Civilization’s progress lies not in acquiring new knowledge, but in aggregating and utilizing the knowledge we already have. When I asked Melcarek how long he spent solving InnoCentive problems, his answer was telling. “If I don’t know what to do after 30 minutes of brainstorming, I give up.”
Lakhani’s findings may be news to people in business and science, fields in which the vogue for specialization has reigned for many decades. But they dovetail neatly with decades of research in economic sociology, echoing a principle sociologists call “the strength of weak ties.”
In 1972 a young sociologist at Harvard named Mark Granovetter crossed the Charles River and asked some 282 professional, technical and managerial workers in Newton, Massachusetts how they’d found their jobs. Not surprisingly, many of Granovetter’s respondents told him they’d used personal contacts, confirmation of the old saw: “It’s not what you know, it’s who you know.” Granovetter, however, dug a little deeper. What kind of personal contact, he wanted to know. Spouse? Brother? Best friend? The majority, it turned out, answered no, no and no. Only 16.7 percent found their job through a close acquaintance, while the rest had secured their position through someone they barely knew. The people who were most helpful were the friends of friends. The people we know well know all the same things—the same eligible singles, the same job openings and the same available apartments—that we know. What Granovetter discovered is that it’s not, in fact, who we know that gets us ahead in life, it’s who we don’t know well.
The strength of weak ties doesn’t just apply to getting jobs. “There’s a strong tendency in human networks to homophily, which means that birds of a feather stick together,” says Lakhani. “And so even when a company chooses to look to external sources for a solution to a problem, they’ll rely on people and companies and labs they already know well, so they run into the same local search biases that are present in internal problem solving.” In this light, it no longer seems so mysterious that the leading chemists at a company like P&G might fail to solve a problem that Ed Melcarek can knock off over a couple glasses of brandy.
The key to making it work is to broadcast the problem using a massive network like InnoCentive’s. Or to return to Lakhani’s metaphor, to make your flower attractive to as many insects as possible. That’s easier said than done. “Firms aren’t set up to broadcast their inner problems to outsiders. Traditional corporate culture is geared to limit outsider access to insider information, not increase it. And what could be more insidery than some problem they’re really stumped on?” Of course, this only makes for greater opportunities for companies willing to swim against the tide.
If an uneducated cabinetmaker can solve one of the most perplexing challenges of his day, if an electrical engineer can solve some of the thorniest chemistry problems encountered by Fortune 500 companies, then maybe MATLAB’s Ned Gulley isn’t off base in suggesting that the collective brain might one day cure cancer. It’s not as remote a prospect as it sounds. Taking a page from the distributed computing project, SETI@home, Stanford University’s chemistry department created Folding@home, which uses the excess capacity of hundreds of thousands of individual PCs to simulate protein folding—the process in which proteins combine to form biological molecules—a crucial step in understanding diseases like cystic fibrosis, Alzheimer’s disease and cancer.
It’s a short leap to go from tapping the excess capacity of thousands of computers to tapping the excess capacity of thousands of brains. This gets to the heart of F.A. Hayek’s proposition: the answers to our present day’s equivalent of the longitude problem lie not in the acquisition of more knowledge, but simply in gathering up all the knowledge that is already spread out among individuals throughout the world. In fact, it’s already occurring. What happens when the Billion—the approximate size of the crowd, which is to say, the number of people connected to the Internet—becomes the Three Billion. What feats of cognition might all these brains, working together, produce?
The possibility of crowdsourcing difficult problems has even been broached in that most unlikely crucible of innovation—the federal government. In October 2007, Senator Bernie Sanders, an independent from Vermont, introduced a bill that would replace the drug monopolies given to pharmaceutical companies by the patent department with a system of cash prizes. The bill calls for the government to create a $80 billion fund, which would then be awarded for narrowly targeted medical objectives, such as developing part of a cure for malaria. At present, pharmaceutical companies have little incentive to conduct research into such life-saving treatments. Drug firms like Merck make an easy target for their seeming callous disregard for killers like malaria, but the fact is that research and development of such drugs is incredibly expensive, and the readiest customers are also the poorest, meaning the pharmaceutical companies would have a very difficult time earning their money back on a malaria drug. Sanders’ bill would address this issue by essentially guaranteeing compensation to any individual or company who successfully develops such a drug.
Sanders isn’t alone in conceiving of problem-solving crowdsourcing as a potential alternative to traditional R&D. Awarding prizes for breakthroughs in medical science have been proposed by former US Senator John Edwards. On the other side of the aisle, Newt Gingrich has proposed a similar system for alleviating government spending. As William Saleton wrote in Slate in October of 2007, Gingrich suggests that “instead of giving $1 billion to a federal agency to deal with a problem … offer the money as a prize to the first company that solves it. As the conversation proceeds, Gingrich throws money at one challenge after another. Hydrogen fuel? Dangle a 10-figure prize.”
The interesting question that all this evokes is "Does crowdsourcing scale?" You argue well for the fact that by spreading out a problem and incentivizing, one can get interesting new answers. But does it scale. Yes, in the case of Linux coding it clearly did so, and the crowd organized itself in interesting ways. But in most of the examples that one puts forward, it is the case of either single designs or single inventions, often then improved by a more traditional R&D-unit. When one guesses that the problem of cancer could be solved through crowdsourcing, one ignores much of the tricky combinatorics that lies behind more complicated systemic innovations. Yes, with enough people behind it, we might be able to map genomic sequences over and over again until we hit upon something, but can all this be automated? Is there no need for a more centralized oversight?
When my laptop runs the SETI@home, it just means that I lend a few cycles, dumb cycles, and that these will be filtered by an algorithm created by the people behind SETI@home, and if I by chance come upon something, the people from S@h will look at it and decide if they've found something. In a sense this would be a very "pure" utilization of resources, effectively doing something with a resource that otherwise would lie fallow.
The logic of crowdsourcing says something similar, but forgets one thing. Those who do not succeed to see their work prominently displayed will, unlike my laptop, care about the fact that they've spent time that didn't really help anyone (least of all themselves).
Scaling crowdsourcing would somehow be able to muster both the greater systemic logic to oversee a project, and at the same time be able to overcome the problem of what happens to the underutilized (and not necessarily replicating) resources. As an example, just look at the number of open source projects that either fail or never get off the ground at all. Yes, I know the counter-argument is that there are enough "cycles" to go around, but I have yet to see any solid evidence for this...
A thorny one, that.
Posted by: Alf Rehn | May 17, 2008 at 09:15 AM
Great read Jeff!
More quibbles.
“The current R&D model is broken.” P&G is one of InnoCentive’s earliest and best customers, but the company works with other crowdsourcing networks as well.”
I wouldn’t have abbreviated Procter & Gamble immediately following R&D.
“Not every quick and curious intellect can land a plum research post at a university or privately funded lab. Some must make HVAC systems.”
Why did you choose HVAC systems as the lowest common denominator against a plum research post at a university or privately funded lab?
“Fortune 500 companies, then maybe MATLAB’s Ned Gulley isn’t off base in suggesting that the collective brain might one day cure cancer.”
Calling it a collective brain Jeff is just spin That is quite a stretch from Collective Intelligence
“But they dovetail neatly with decades of research in economic sociology, echoing a principle sociologists call “the strength of weak ties.”
Very interesting concept, the fact that it is important to recognize the difference between a nod tie, weak tie and an absent one gives it some context Jeff.
Cheers, Alan
Posted by: alan booker | May 17, 2008 at 02:58 PM
Just a quick note to future commenters: Alan's approach here is spot on, in that he quotes the relevant passage he's responding to, then writes his comment below. The idea is to create end-note symbols that will correspond to the appropriate parts of the appendix. If you reply generally to the excerpt, it makes this task more difficult (though not impossible.) Make sense?
Posted by: Jeff Howe | May 17, 2008 at 03:09 PM
First, thanks for sharing all this on your blog! I'll start with a tiny copy-editing quibble. In this passage:
“I read that and thought, ‘Huh. Kind of like open source. Someone poses a problem and all sorts of random, strange people show up and say I’ve got an answer for you, and it’s never the one you’d anticipate.”
You need an end-quote for the thought beginning with 'Huh.'
Now for the deeper thought:
"The longitude prize constitutes the earliest known example of crowdcasting—broadcasting a problem to the widest possible audience in the blind hope that someone, somewhere—maybe even a Yorkshire cabinetmaker—will come up with a solution."
I got here through a little crowdcasting of your own -- I saw a post on Tish Grier's blog indicating you needed a little more involvement in this blog. That suggests to me that crowdsourcing *can* scale, but that it takes a lot of work to make it do so. It sounds like you'd like the editing process of your book on crowdsourcing to be crowdsourced itself -- a laudable goal -- but it's not as easy as it sounds, is it? Good crowdsourcing needs to achieve critical mass somehow, through word-of-mouth (word-of-blog?) or some other mechanism. In short, good old-fashioned publicity. And so in its own way, the new is old.
This is just my random riff on a quick read of this chapter. I may have more to say after I read more of the book. I just got here, so please forgive me if other commenters have already said much the same thing.
Posted by: Mike Ho | May 19, 2008 at 09:08 AM