How do we know what we think we know? To narrow this longstanding epistemological question, let me ask this about the world I generally inhabit—medicine, where I work as a neuroscientist. For questions about medical conditions, two sources of knowledge exist. There is expert knowledge—the kind acquired by those who read the primary scientific papers, examine findings from controlled studies, and who, by virtue of their training and their advanced degrees, carry the weight of authority. The second is what today would be called “wiki” knowledge, the kind that arises from collective experience. Today, the knowledge of the designated expert is increasingly challenged by the collective experience of ever-expanding cybercommunities. In the battle of the blogosphere vs. the expert, the expert seems to be losing ground. This contemporary dialectic represents a challenge for many disciplines, including the journalist, who must decide how to balance expert views with those of the cybercommunity.
Knowledge: Expert Vs. Wiki
When medical findings are announced, whether a new therapy, a new preventive measure, or a new research finding, neither the journalist nor the physician should assume that an expert opinion is definitive. The expert may be “as good as it gets,” but the limitations of the expert approach need to be clear. For example, let’s take treatment decisions with a newly approved medication for Alzheimer’s disease. To get approved by the FDA, the pharmaceutical company had to prove safety and efficacy. But how frequently does the drug fail to work, and do other health-related factors such as lifestyle or coexisting disease or genetic risk affect the likelihood the drug will work? These are difficult questions for the expert. In the case of the most commonly used drug in Alzheimer’s disease—donepezil—the physician has no idea about enhanced or diminished benefit in association with other health factors and usually does not mention to the family that many users show no benefit at all.
Perhaps the power of the wiki could provide more depth when one is making a decision about a drug treatment. Certainly, the choice of a medication becomes even more acute for some of the stratospherically priced drugs used in cancer treatment today. So how can we create a wiki-based knowledge environment for medical information? In times past, collective knowledge derived from folk medicine, old wives tales, and anecdotal reports. The number of contributors to collective knowledge in any one community was small and, therefore, the conclusions clinically suspect.
The modern-day version of folk medicine is no longer confined to a small circle of happenstance encounters within the limits of our physical geography. With the disappearance of these boundaries, our links to medical conditions like our own can reach across the globe. Large numbers of people—well beyond the numbers found in most medical studies—can build disease-oriented social networks with layers of added information and with an ease of follow-up to create a living, dynamic wiki. From the network one can cluster individuals in any way desired—by geographic location, by occupation, by response to a medication—and begin to extract patterns and correlations. We can organize and reorganize data and perform statistics based on any parameter we chose and create hypotheses that can then be verified prospectively.
Within the potential of social networks lies untapped wiki knowledge poised to challenge the experts by opening wide the collective knowledge gate. In November, Google announced its new Web tool—Google Flu Trends—which uses people’s search clues (entering phrases such as “flu symptoms”) to create graphs and maps to predict and show regional outbreaks of the flu.
Can social networks rival what is learned from expert approaches such as controlled studies and disease registries? Sound conclusions in the medical field are based upon statistical significance. The statistical power of a population, i.e. the ability to distinguish between an experimental and control group, when posed a research question often depends on having a sufficiently large study group. The best way to increase the number of participating individuals is tapping into the Internet. However, saddled with a freewheeling Wild West style, the Internet cannot easily provide pure well-controlled study populations. But the vast potential for touching enormous numbers of people could negate the noise of the Web. Experts use “meta-analysis” to increase the size of their experimental sample. Wiki knowledge derived from a social network offers a fluid, open source, ongoing meta-analysis—a virtual collection of experiences that can be constantly updated as users enter more individual data.
Benefits and Challenges of Collective Information
Social networks empower the “expert,” be it a doctor or a journalist, because access to this community-generated knowledge is shared by all. For example, illness and a significant story intersected at Love Canal, where 21,000 tons of chemical waste lay buried beneath the community unbeknownst to the residents. Back in 1978, a time long before social media existed, Lois Gibbs, a local mother and president of the Love Canal Homeowners’ Association, first associated exposure to the leaking chemical waste with the epilepsy, asthma and urinary tract infections that were recurring in her children. Although flagrant and clear cut, Love Canal is not unique. Now, the ability of Web-based medical networks to cluster data geographically has the potential to reveal other dangerous living conditions. Similarly, occupational risks for disease are well recognized, and organizing medical data in this way will likely serve as an early warning system for on-the-job risks—and for investigative stories that can be done about them.
Figuring out what constitutes a healthy lifestyle is something that consumes the time of both doctors and journalists, whose job it is to report reliably on the barrage of evidence emerging from many different studies—much of it contradictory or, at least, confusing. New information surfaces almost daily about dietary measures or fitness programs that will increase or decrease our risk for cancer, heart attacks, Alzheimer’s disease, and more, and some of it is potentially contaminated by the bias of financial involvement.
How can one possibly capture all these simultaneous variables when computing risk? Did a study that found something new about coffee drinkers control for the number of hours those people spent in the gym? When a new drug is tested, the control group may not be identical to the experimental group in caloric intake, number of portions of vegetables eaten, or their amount of daily exercise. At best, the study is controlled for age and gender. But if those taking the drug are eating poorly and under stress and those on placebo are dining on salads and jogging on the beach, the wrong conclusion could be reached. And if one adds in genetic variation found in human populations—certain types of genes can increase or diminish the risk for disease—the variables mount further. Controlled studies are just not powered to capture multiple variables, and medical conditions are brimming with variables. The only way to increase the statistical power of a conclusion is to increase the sample size, exactly what social networks are designed to do.
Because wiki knowledge in the social network arena is obtained in an unconventional manner, it might not provide conclusive evidence. Therefore, a preferable way of thinking about wiki knowledge is as a guidepost for the design of hypotheses (for scientists to test) or generating story ideas (for journalists to report). For each of us, the pitfalls are evident, and a few of them are highlighted below:
- Selection bias is a problem. Those on a social network tend to be younger and not economically disadvantaged. When groups of people are excluded due to entry barriers, the information generated from the community will be biased, and other knowledge will be lost or skewed. In time, the increasing penetration of the Internet to all segments of the society will resolve this issue as has happened for telephones and TV.
- The privacy question. No network is totally secure—and medical information is not immune to the problem. This summer, staff at a hospital near Los Angeles was discovered snooping through records of Hollywood celebrities. And this case is not unique. Beyond the security of servers, networks allow levels of access; therefore, on a site where people share medical information, they can limit the information that others can see. Some individuals may want to remain completely anonymous. Others may be willing to share all their information within a small subnetwork of people they know well and keep anonymous their data to the larger network.
- Entry of false data is a potentially serious issue—for doctors and journalists alike. For example, take reporting on the performance of surgeons, an area in which data are sorely needed. Suppose a disgruntled patient wants to smear a surgeon and fabricates multiple entries with bad outcomes. Tools are needed for verification. Suppose a person is part of social network related to weight control or hypertension and enters false data. If just a few people are guilty of false entries the overall conclusions will not vary much. But large numbers of people may have a tendency to lie or distort their personnel information even if their identity is concealed.
“Net Geners Relate to News in New Ways”
– Don TapscottNeither doctors nor journalists will have been the first to venture into the realm of figuring out how to utilize wiki knowledge. In “Wikinomics,” by Don Tapscott and Anthony Williams, many positive examples are presented that bring collective Web-based knowledge to the business model. Yet there are critics of this approach, too. Andrew Keen, author of “The Cult of the Amateur,” argues that Web-based knowledge is superficial and lacks deep and considered judgment. Indeed, Web content can be boisterous, unfiltered and amateur. Yet if conventional knowledge is only what experts know, then should everything else be considered “amateur” knowledge? While Web users may be a raucous bunch, they can be as easily airbrushed into a statistic through a social network as they can be in an expert study.
An External Hard Drive for the Brain
As a neuroscientist who spends time thinking about how people’s brains process information, this technology—and the information overflow it brings—are without a doubt changing the way human beings make decisions. Neuroscientists have increasingly come to understand memory as a function not intended to recreate the past, but to guide us into the future. Viewed in this way, memory does not have to be perfectly accurate; instead it has to serve us for outcome simulation and correct decision-making based on the memory of an experience that resembled our current circumstances.
Stores of information downloaded from hand-held devices will help close the gap between successful and unsuccessful outcomes when making decisions because we can draw upon a deep base of information and experience. We can instantly tap into a living source of collective experience about our condition while sitting with the physician. As pointed out by Daniel L. Schacter and Donna Rose Addis in a recent essay in Nature (2007), “information about the past is useful only to the extent that it allows us to anticipate what may happen in the future.” Our ability to anticipate the future may be enhanced by a richer store of information that includes a Web-based compilation of data.
It is perhaps an irony of our time that with all of these avenues to discover knowledge at our command, we can find ourselves starved for information in a sea churning with nothing but information. The particular knowledge craved, for example, by those given a life-threatening diagnosis, often lies outside the expertise of physicians—even specialists. While flickers of hope appear on the Web through encounters with others and a shared experience, judging the reliability of this experience—and its fit with our own—can be difficult. But to have the opportunity to find information and test its reliability means that no longer is one person—an expert—expected to know everything and render infallible judgment. That view is the no-longer tenable burden of the expert physician; nor can it any longer be the guiding belief of the trained journalist.
Kenneth S. Kosik is the Harriman Professor of Neuroscience Research and codirects the Neuroscience Research Institute at the University of California, Santa Barbara. Kosik is a founder of the Learning & the Brain conference based in Boston and the founder of the center for Cognitive Fitness and Innovative Therapies in Santa Barbara.