A large body of evidence supports the concept that groups of people make better decisions than individuals. This idea has become known as the wisdom of the crowd. The power of crowds to make accurate predictions has been recognized for almost 100 years in the form of prediction markets. Prediction markets have that name because that is precisely what they do—make predictions. In its most common form, a prediction market is incredibly simple. The market creates a security (takes a bet) that a future event will occur. The security pays $1 if the event takes place, $0 if it
does not. Through the wisdom of the crowd, the price of the security reflects the market’s assessment of the probability of that future event.

For example, a prediction market security might pay $1 if the New England Patriots win the upcoming Super Bowl and pay nothing if they do not. The Super Bowl occurs each year, and the winner is determined with absolute certainty. In this example, the price of the security during the NFL regular season is the collective assessment of the investing (betting) crowd that the Patriots will win. A price of $1 means that victory in the Super Bowl is guaranteed, and a price of $0 means that winning is impossible, for example, by being eliminated from the playoffs. Any price between $0 and $1 reflects the market’s current assessment of the likelihood of a Super Bowl win by the Patriots.

These sorts of crowd-sourced markets have been used to predict outcomes of events that are highly complex, even those that are fluid day-to-day. Prediction markets have been used to predict outcomes of political elections since the presidential race between Charles Evans Hughes and Woodrow Wilson in 1916. The Iowa Electronic Market is a well-known research prediction market, in existence since the 1988 presidential election. Figure 1.1 demonstrates the likelihood that the 2008 presidential election would be won by the Democratic or the Republican candidate, as forecast by the Iowa Electronic Market.

Prediction markets reflect the power of one form of cognitive diversity. Accurate forecasting occurs because very large numbers of individuals, with correspondingly very diverse perspectives, are actively engaged in solving a complex problem related to a future event. Predictive power occurs when individuals with varied perspectives consider a common set of facts. In the example of the 2008 election, millions of individuals were aware of the domestic and international events depicted in the timeline of Figure 1.1 (and many other events not noted). Each individual (investor)
received the same information. Each judged the events as relevant or irrelevant to the upcoming election, important or trivial, of lasting importance or transient. Each applied a different perspective based upon his or her own personal experiences in making these judgments. No two individuals have identical life experiences, and no two are likely to process these complex inputs identically or to predict the impending election results identically. Yet the collective assessment of the crowd proved to be remarkably prescient. Predictive power comes from differences, not similarities.

Predictive markets have lessons to teach that are relevant to the practice of surgery. A multidisciplinary tumor board convenes a much smaller group of decision-makers and provides treatment recommendations in the moment. Optimizing clinical rotations for a surgical residency, crucial for the involved trainees, is performed by a committee of faculty members and house staff. While it is not possible to enlist thousands (millions) of decision-makers, intentional efforts to welcome differing perspectives can lead to more accurate treatment or teaching recommendations.

The reordering of scientific discovery from an area of individual brilliance to a team endeavor over the past several decades is striking. Large-scale collaborative projects have become increasingly common in fields ranging from medicine to high-energy physics, and entirely new fields of science have developed as a result, with genomics and bioinformatics as examples.

The rise of larger scientific teams can be quantified using authorship of scientific papers. The number of co-authors per paper has increased steadily for most of the past century, growing by 25% from 1997 through 2016.⁷ The evolution of team science has also increased scientific impact. When primary manuscripts are considered, increased co-authorship is associated with an increased citation index. The positive effect of co-author number on citations has been demonstrated for both biomedical and physical sciences. A recent study mapped collaborative scientific dynamics on a mass scale. Parish, Boyack, and Ioannidis examined 249,054 investigators using the Scopus database (https://www.scopus.com/) for the period 2006 to 2015 in order to construct a collaboration index.⁸ In medicine and health sciences, greater collaboration was associated with a higher citation index (h-index).

When individual investigators are examined, those with larger numbers of collaborators demonstrate greater scientific productivity. Recent studies have ascribed the positive effects of co-authorship to the establishment of collaboration networks. Collaboration networks represent scientific communities with investigators connected by ideas (Figure 1.2). Positive network effects have been demonstrated in fields ranging from statistics to pharmacology and informatics. Increased network effects are associated with greater scientific productivity. Not surprisingly, the correlations between collaboration, networking, and scientific productivity increase as investigators progress professionally as more time spent as part of a scientific team permits relationships to persist and deepen. International collaborations are also very strongly correlated with scientific productivity.

Recent evidence suggests that networks that provide greater ethnic diversity offer competitive advantage, allowing teams to publish more frequently cited papers. Freeman and Huang used bibliometric techniques to examine 2.5 million papers, all written by American authors.⁹ Ethnicity was implied by analysis of surnames. For example, Smith was assigned English ethnicity; Chang was assigned Chinese ethnicity. Manuscripts with four or more authors in which authors had multiple ethnicities were cited more frequently than those written by authors of the same ethnicity.
The advantage of ethnic diversity represented a difference of 5% to 10% additional citations per publication. Ethnic homogeneity was associated with publication in lower-impact journals and fewer citations. Freeman and Huang believe that more ethnically diverse research groups benefit from different, richer networks and perhaps a greater variety of perspectives.

The effects of sex and gender diversity are also emerging in science. Female authorship in high-impact medical journals increased from 27% to 37% in the years 1994 to 2014.¹⁰ Biological sex and gender (behaviors and social attitudes associated with being a man or woman) are crucial factors in human health. Biological sex and gender also clearly shape life experiences, with effects on cognition and perspective that are relevant to the scientific workplace.

In spite of direct relevance, the effects of biological sex and gender on disease risk and health outcomes have been widely neglected in the medical literature. Attention to gender and sex in clinical trial design is an essential element of scientific quality. In a recent study, a global sample of more than 1.5 million medical research publications examined the effects of women’s authorship on gender and sex analysis as a means of improving scientific validity of health studies.¹¹ Women’s participation as first or last author was strongly correlated with use of gender and sex analysis. The effect attributable to women’s participation was strongest when women served as senior authors.

The benefits of scientific teams are also apparent in corporate research and development. Reagans and Zuckerman analyzed 224 research and development teams in 29 corporations across seven industries: automotive, chemicals, electronics, aerospace, pharmaceuticals, biotechnology, and energy.¹² Some teams were productive, others less so. Teams that demonstrated both demographic heterogeneity and strong communication exhibited enhanced learning capabilities and high productivity. Homogeneous teams or those that were unable to develop strong connections among team members were underproductive. Greater demographic diversity brings together people with different skills, information, life experiences, and perspectives. These attributes enhance creative problem solving if team members allocate time to activities that cross the bounds of individual expertise. However, crossing boundaries is inherently uncomfortable and requires trust and strong communication.

While scientific collaboration is unquestionably more prevalent today than in past years, not all scientific teams are equally productive. The most successful teams are characterized by high levels of integration and interaction.

Recent research has examined the elements of productive scientific teams. The foundation of a successful team is a compelling scientific question. To address the question, the productive team brings together members with specific, different, and complementary scientific expertise. Regular team meetings are crucial. Scheduled meetings are obviously important to discuss the overall scientific goal, to highlight the different objectives of the individuals composing the team, to share and critique data, and to plan next steps. Less obvious but more important, regular meetings serve a socialization purpose, helping team members form bonds with each other. These bonds promote trust, which is the single most important factor in team science. Trust permits scientific teams to engage in disagreement while limiting conflict. Trust enables the equitable sharing of credit and authorship. Often, successful scientific teams have an overall lead, but other members have key leadership roles that are relevant to achieving the overall goal. Leadership is participatory.

The National Cancer Institute has examined the science of team science. The major influences on successful team science are not technological, rather interpersonal and organizational. Stokols et al. cite interpersonal factors to include team members’ familiarity, informality, and social cohesiveness; diversity of members’ perspectives and abilities; regular and effective communication among members to develop consensus and shared goals; and establishment of a hospitable conversational environment through mutual respect among team members.¹³ Organizations that are nonhierarchical facilitate autonomy and participatory goal setting. Spatial proximity allows frequent opportunities for face-to-face communication and informal information exchange. To be sure, technological resources are important. But social factors trump technological prowess every time.