Search

A large-scale study provides a causal test for a cornerstone of social science

Science’s changing demographics raise new questions about research team diversity and research outcomes. We study mixed-gender research teams, examining 6.6 million papers published across the medical sciences since 2000 and establishing several core findings. First, the fraction of publications by mixed-gender teams has grown rapidly, yet mixed-gender teams continue to be underrepresented compared to the expectations of a null model. Second, despite their underrepresentation, the publications of mixed-gender teams are substantially more novel and impactful than the publications of same-gender teams of equivalent size. Third, the greater the gender balance on a team, the better the team scores on these performance measures. Fourth, these patterns generalize across medical subfields. Finally, the novelty and impact advantages seen with mixed-gender teams persist when considering numerous controls and potential related features, including fixed effects for the individual researchers, team structures, and network positioning, suggesting that a team’s gender balance is an underrecognized yet powerful correlate of novel and impactful scientific discoveries.

Significance Mentorship is arguably a scientist’s most significant collaborative relationship; yet of all collaborations, comparatively little research exists on the link between mentorship and protégé success. Using new large-scale data from the genealogical and performance records of 10s of thousands of scientists worldwide from 1960 to the present, we found that mentorship is associated with diverse forms of protégé success, significantly increasing protégés’ chances of producing celebrated research, being inducted into the National Academy of Science, and achieving superstardom. Paradoxically, protégés achieve their highest impact when they display intellectual independence from their mentors. Protégés do their best work when they break from their mentor’s research topics and coauthor no more than a small portion of their overall research with their mentors., Einstein believed that mentors are especially influential in a protégé’s intellectual development, yet the link between mentorship and protégé success remains a mystery. We marshaled genealogical data on nearly 40,000 scientists who published 1,167,518 papers in biomedicine, chemistry, math, or physics between 1960 and 2017 to investigate the relationship between mentorship and protégé achievement. In our data, we find groupings of mentors with similar records and reputations who attracted protégés of similar talents and expected levels of professional success. However, each grouping has an exception: One mentor has an additional hidden capability that can be mentored to their protégés. They display skill in creating and communicating prizewinning research. Because the mentor’s ability for creating and communicating celebrated research existed before the prize’s conferment, protégés of future prizewinning mentors can be uniquely exposed to mentorship for conducting celebrated research. Our models explain 34–44% of the variance in protégé success and reveals three main findings. First, mentorship strongly predicts protégé success across diverse disciplines. Mentorship is associated with a 2×-to-4× rise in a protégé’s likelihood of prizewinning, National Academy of Science (NAS) induction, or superstardom relative to matched protégés. Second, mentorship is significantly associated with an increase in the probability of protégés pioneering their own research topics and being midcareer late bloomers. Third, contrary to conventional thought, protégés do not succeed most by following their mentors’ research topics but by studying original topics and coauthoring no more than a small fraction of papers with their mentors.

Replicability tests of scientific papers show that the majority of papers fail replication. Moreover, failed papers circulate through the literature as quickly as replicating papers. This dynamic weakens the literature, raises research costs, and demonstrates the need for new approaches for estimating a study’s replicability. Here, we trained an artificial intelligence model to estimate a paper’s replicability using ground truth data on studies that had passed or failed manual replication tests, and then tested the model’s generalizability on an extensive set of out-of-sample studies. The model predicts replicability better than the base rate of reviewers and comparably as well as prediction markets, the best present-day method for predicting replicability. In out-of-sample tests on manually replicated papers from diverse disciplines and methods, the model had strong accuracy levels of 0.65 to 0.78. Exploring the reasons behind the model’s predictions, we found no evidence for bias based on topics, journals, disciplines, base rates of failure, persuasion words, or novelty words like “remarkable” or “unexpected.” We did find that the model’s accuracy is higher when trained on a paper’s text rather than its reported statistics and that n-grams, higher order word combinations that humans have difficulty processing, correlate with replication. We discuss how combining human and machine intelligence can raise confidence in research, provide research self-assessment techniques, and create methods that are scalable and efficient enough to review the ever-growing numbers of publications—a task that entails extensive human resources to accomplish with prediction markets and manual replication alone.

Polarization affects many forms of social organization. A key issue focuses on which affective relationships are prone to change and how their change relates to performance. In this study, we analyze a financial institutional over a two-year period that employed 66 day traders, focusing on links between changes in affective relations and trading performance. Traders’ affective relations were inferred from their IMs (>2 million messages) and trading performance was measured from profit and loss statements (>1 million trades). Here, we find that triads of relationships, the building blocks of larger social structures, have a propensity towards affective balance, but one unbalanced configuration resists change. Further, balance is positively related to performance. Traders with balanced networks have the “hot hand”, showing streaks of high performance. Research implications focus on how changes in polarization relate to performance and polarized states can depolarize., How do socially polarized systems change and how does a change in polarization relate to performance? Using instant messaging data and performance records from day traders, the authors find that certain relations are prone to balance and that balance is associated with better trading decisions.

Fast growing scientific topics have famously been key harbingers of the new frontiers of science, yet, large-scale analyses of their genesis and impact are rare. We investigated one possible factor connected with a topic’s extraordinary growth: scientific prizes. Our longitudinal analysis of nearly all recognized prizes worldwide and over 11,000 scientific topics from 19 disciplines indicates that topics associated with a scientific prize experience extraordinary growth in productivity, impact, and new entrants. Relative to matched non-prizewinning topics, prizewinning topics produce 40% more papers and 33% more citations, retain 55% more scientists, and gain 37 and 47% more new entrants and star scientists, respectively, in the first five-to-ten years after the prize. Funding do not account for a prizewinning topic’s growth. Rather, growth is positively related to the degree to which the prize is discipline-specific, conferred for recent research, or has prize money. These findings reveal new dynamics behind scientific innovation and investment., Scientific revolutions have famously inspired scientists and innovation but large-scale analyses of scientific revolutions in modern science are rare. Here, the authors investigate one possible factor connected with a topic’s extraordinary growth—scientific prizes.

As terror groups proliferate and grow in sophistication, a major international concern is the development of scientific methods that explain and predict insurgent violence. Approaches to estimating a group’s future lethality often require data on the group’s capabilities and resources, but by the nature of the phenomenon, these data are intentionally concealed by the organizations themselves via encryption, the dark web, back-channel financing, and misinformation. Here, we present a statistical model for estimating a terror group’s future lethality using latent-variable modeling techniques to infer a group’s intrinsic capabilities and resources for inflicting harm. The analysis introduces 2 explanatory variables that are strong predictors of lethality and raise the overall explained variance when added to existing models. The explanatory variables generate a unique early-warning signal of an individual group’s future lethality based on just a few of its first attacks. Relying on the first 10 to 20 attacks or the first 10 to 20% of a group’s lifetime behavior, our model explains about 60% of the variance in a group’s future lethality as would be explained by a group’s complete lifetime data. The model’s robustness is evaluated with out-of-sample testing and simulations. The findings’ theoretical and pragmatic implications for the science of human conflict are discussed.

Climb atop shoulders and wait for funerals. That, suggested Newton and then Planck, is how science advances (more or less). We've come far since then, but many notions about how people and practices, policies, and resources influence the course of science are still more rooted in traditions and intuitions than in evidence. We can and must do better, lest we resign ourselves to “intuition-based policy” when making decisions and investments aimed at driving scientific progress. Science invited experts to highlight key aspects of the scientific enterprise that are steadily yielding to empirical investigation—and to explain how Newton and Planck got it right (and Einstein got it wrong). — Brad Wible

Scientific prizes are among the greatest recognition a scientist receives from their peers and arguably shape the direction of a field by conferring credibility to persons, ideas, and disciplines, providing financial rewards, and promoting rituals that reinforce scientific communities. The proliferation of prizes and links among prizes suggest that the prize network embodies information about scientists and ideas poised to grow in acclaim. Using comprehensive new data on prizes and prizewinners worldwide and across disciplines, we examine the growth dynamics and interlocking relationships found in the worldwide scientific prize network. We focus on understanding how the knowledge linkages among prizes and scientists' propensities for prizewinning are related to knowledge pathways across disciplines and stratification within disciplines. We find several key links between prizes and scientific advances. First, despite a proliferation of diverse prizes over time and across the globe, prizes are more concentrated within a relatively small group of scientific elites, and ties within the elites are more clustered, suggesting that a relatively constrained number of ideas and scholars lead science. Second, we find that certain prizes are strongly interlocked within and between disciplines by scientists who win multiple prizes, revealing the key pathways by which knowledge systematically gains credit and spreads through the network. Third, we find that genealogical and co authorship networks strongly predict who wins one or more prizes and explains the high level of interconnections among acclaimed scientists and their path breaking ideas.

BACKGROUND The increasing availability of digital data on scholarly inputs and outputs—from research funding, productivity, and collaboration to paper citations and scientist mobility—offers unprecedented opportunities to explore the structure and evolution of science. The science of science (SciSci) offers a quantitative understanding of the interactions among scientific agents across diverse geographic and temporal scales: It provides insights into the conditions underlying creativity and the genesis of scientific discovery, with the ultimate goal of developing tools and policies that have the potential to accelerate science. In the past decade, SciSci has benefited from an influx of natural, computational, and social scientists who together have developed big data–based capabilities for empirical analysis and generative modeling that capture the unfolding of science, its institutions, and its workforce. The value proposition of SciSci is that with a deeper understanding of the factors that drive successful science, we can more effectively address environmental, societal, and technological problems. ADVANCES Science can be described as a complex, self-organizing, and evolving network of scholars, projects, papers, and ideas. This representation has unveiled patterns characterizing the emergence of new scientific fields through the study of collaboration networks and the path of impactful discoveries through the study of citation networks. Microscopic models have traced the dynamics of citation accumulation, allowing us to predict the future impact of individual papers. SciSci has revealed choices and trade-offs that scientists face as they advance both their own careers and the scientific horizon. For example, measurements indicate that scholars are risk-averse, preferring to study topics related to their current expertise, which constrains the potential of future discoveries. Those willing to break this pattern engage in riskier careers but become more likely to make major breakthroughs. Overall, the highest-impact science is grounded in conventional combinations of prior work but features unusual combinations. Last, as the locus of research is shifting into teams, SciSci is increasingly focused on the impact of team research, finding that small teams tend to disrupt science and technology with new ideas drawing on older and less prevalent ones. In contrast, large teams tend to develop recent, popular ideas, obtaining high, but often short-lived, impact. OUTLOOK SciSci offers a deep quantitative understanding of the relational structure between scientists, institutions, and ideas because it facilitates the identification of fundamental mechanisms responsible for scientific discovery. These interdisciplinary data-driven efforts complement contributions from related fields such as scientometrics and the economics and sociology of science. Although SciSci seeks long-standing universal laws and mechanisms that apply across various fields of science, a fundamental challenge going forward is accounting for undeniable differences in culture, habits, and preferences between different fields and countries. This variation makes some cross-domain insights difficult to appreciate and associated science policies difficult to implement. The differences among the questions, data, and skills specific to each discipline suggest that further insights can be gained from domain-specific SciSci studies, which model and identify opportunities adapted to the needs of individual research fields.

Peer-to-peer lending is hypothesized to help equalize economic opportunities for the world's poor. We empirically investigate the "flat-world" hypothesis, the idea that globalization eventually leads to economic equality, using crowdfinancing data for over 660,000 loans in 220 nations and territories made between 2005 and 2013. Contrary to the flat-world hypothesis, we find that peer-to-peer lending networks are moving away from flatness. Furthermore, decreasing flatness is strongly associated with multiple variables: relatively stable patterns in the difference in the per capita GDP between borrowing and lending nations, ongoing migration flows from borrowing to lending nations worldwide, and the existence of a tie as a historic colonial. Our regression analysis also indicates a spatial preference in lending for geographically proximal borrowers. To estimate the robustness for these patterns for future changes, we construct a network of borrower and lending nations based on the observed data. Then, to perturb the network, we stochastically simulate policy and event shocks (e.g., erecting walls) or regulatory shocks (e.g., Brexit). The simulations project a drift towards rather than away from flatness. However, levels of flatness persist only for randomly distributed shocks. By contrast, loss of the top borrowing nations produces more flatness, not less, indicating how the welfare of the overall system is tied to a few distinctive and critical country-pair relationships.

A new analysis of biomedical awards over five decades shows men receive more cash and more respect for their research than women do, report Brian Uzzi and colleagues. A new analysis of biomedical awards over five decades shows men receive more cash and more respect for their research than women do, report Brian Uzzi and colleagues.

Many leaders today do not rise through the ranks but are recruited directly out of graduate programs into leadership positions. We use a quasi-experiment and instrumental-variable regression to understand the link between students' graduate school social networks and placement into leadership positions of varying levels of authority. Our data measure students' personal characteristics and academic performance, as well as their social network information drawn from 4.5 million email correspondences among hundreds of students who were placed directly into leadership positions. After controlling for students' personal characteristics, work experience, and academic performance, we find that students' social networks strongly predict placement into leadership positions. For males, the higher a male student's centrality in the school-wide network, the higher his leadership-job placement will be. Men with network centrality in the top quartile have an expected job placement level that is 1.5 times greater than men in the bottom quartile of centrality. While centrality also predicts women's placement, high-placing women students have one thing more: an inner circle of predominantly female contacts who are connected to many nonoverlapping third-party contacts. Women with a network centrality in the top quartile and a female-dominated inner circle have an expected job placement level that is 2.5 times greater than women with low centrality and a male-dominated inner circle. Women who have networks that resemble those of high-placing men are low-placing, despite having leadership qualifications comparable to high-placing women.

Making an Impact How big a role do unconventional combinations of existing knowledge play in the impact of a scientific paper? To examine this question, Uzzi et al. (p. 468 ) studied 17.9 million research articles across five decades of the Web of Science, the largest repository of scientific research. Scientific work typically appeared to draw on highly conventional, familiar mixtures of knowledge. The highest-impact papers were not the ones that had the greatest novelty, but had a combination of novelty and otherwise conventional combinations of prior work.

Papers or patents that cite past work of a particular age distribution double their chances of being a hit., Scientists and inventors can draw on an ever-expanding literature for the building blocks of tomorrow’s ideas, yet little is known about how combinations of past work are related to future discoveries. Our analysis parameterizes the age distribution of a work’s references and revealed three links between the age of prior knowledge and hit papers and patents. First, works that cite literature with a low mean age and high age variance are in a citation “hotspot”; these works double their likelihood of being in the top 5% or better of citations. Second, the hotspot is nearly universal in all branches of science and technology and is increasingly predictive of a work’s future citation impact. Third, a scientist or inventor is significantly more likely to write a paper in the hotspot when they are coauthoring than whey they are working alone. Our findings are based on all 28,426,345 scientific papers in the Web of Science, 1945–2013, and all 5,382,833 U.S. patents, 1950–2010, and reveal new antecedents of high-impact science and the link between prior literature and tomorrow’s breakthrough ideas.

Emotions are increasingly inferred linguistically from online data with a goal of predicting off-line behavior. Yet, it is unknown whether emotions inferred linguistically from online communications correlate with actual changes in off-line activity. We analyzed all 886,000 trading decisions and 1,234,822 instant messages of 30 professional day traders over a continuous 2 year period. Linguistically inferring the traders’ emotional states from instant messages, we find that emotions expressed in online communications reflect the same distributions of emotions found in controlled experiments done on traders. Further, we find that expressed online emotions predict the profitability of actual trading behavior. Relative to their baselines, traders who expressed little emotion or traders that expressed high levels of emotion made relatively unprofitable trades. Conversely, traders expressing moderate levels of emotional activation made relatively profitable trades.

Users online tend to select information that support and adhere their beliefs, and to form polarized groups sharing the same view - e.g. echo chambers. Algorithms for content promotion may favour this phenomenon, by accounting for users preferences and thus limiting the exposure to unsolicited contents. To shade light on this question, we perform a comparative study on how same contents (videos) are consumed on different online social media - i.e. Facebook and YouTube - over a sample of 12M of users. Our findings show that content drives the emergence of echo chambers on both platforms. Moreover, we show that the users' commenting patterns are accurate predictors for the formation of echo-chambers.

In theoretical ecology, simple stochastic models that satisfy two basic conditions about the distribution of niche values and feeding ranges have proved successful in reproducing the overall structural properties of real food webs, using species richness and connectance as the only input parameters1, 2, 3, 4. Recently, more detailed models have incorporated higher levels of constraint in order to reproduce the actual links observed in real food webs5, 6. Here, building on previous stochastic models of consumer–resource interactions between species1, 2, 3, we propose a highly parsimonious model that can reproduce the overall bipartite structure of cooperative partner–partner interactions, as exemplified by plant–animal mutualistic networks7. Our stochastic model of bipartite cooperation uses simple specialization and interaction rules, and only requires three empirical input parameters. We test the bipartite cooperation model on ten large pollination data sets that have been compiled in the literature, and find that it successfully replicates the degree distribution, nestedness and modularity of the empirical networks. These properties are regarded as key to understanding cooperation in mutualistic networks8, 9, 10. We also apply our model to an extensive data set of two classes of company engaged in joint production in the garment industry. Using the same metrics, we find that the network of manufacturer–contractor interactions exhibits similar structural patterns to plant–animal pollination networks. This surprising correspondence between ecological and organizational networks suggests that the simple rules of cooperation that generate bipartite networks may be generic, and could prove relevant in many different domains, ranging from biological systems to human society11, 12, 13, 14.

This paper demonstrates that teamwork in science increasingly spans university boundaries, a dramatic shift in knowledge production that generalizes across virtually all fields of science, engineering, and social science. Moreover, elite universities play a dominant role in this shift. By examining 4.2 million papers published over three decades, we found that multi-university collaborations (i) are the fastest growing type of authorship structure, (ii) produce the highest-impact papers when they include a top-tier university, and (iii) are increasingly stratified by in-group university rank. Despite the rising frequency of research that crosses university boundaries, the intensification of social stratification in multi-university collaborations suggests a concentration of the production of scientific knowledge in fewer rather than more centers of high-impact science.

Mechanisms that enable declining networks to avert structural collapse and performance degradation are not well understood. This knowledge gap reflects a shortage of data on declining networks and an emphasis on models of network growth. Analyzing >700,000 transactions between firms in the New York garment industry over 19 years, we tracked this network's decline and measured how its topology and global performance evolved. We find that favoring asymmetric (disassortative) links is key to preserving the topology and functionality of the declining network. Based on our findings, we tested a model of network decline that combines an asymmetric disassembly process for contraction with a preferential attachment process for regrowth. Our simulation results indicate that the model can explain robustness under decline even if the total population of nodes contracts by more than an order of magnitude, in line with our observations for the empirical network. These findings suggest that disassembly mechanisms are not simply assembly mechanisms in reverse and that our model is relevant to understanding the process of decline and collapse in a broad range of biological, technological, and financial networks.

We have used 19.9 million papers over 5 decades and 2.1 million patents to demonstrate that teams increasingly dominate solo authors in the production of knowledge. Research is increasingly done in teams across nearly all fields. Teams typically produce more frequently cited research than individuals do, and this advantage has been increasing over time. Teams now also produce the exceptionally high-impact research, even where that distinction was once the domain of solo authors. These results are detailed for sciences and engineering, social sciences, arts and humanities, and patents, suggesting that the process of knowledge creation has fundamentally changed.

Agents in creative enterprises are embedded in networks that inspire, support, and evaluate their work. Here, we investigate how the mechanisms by which creative teams self-assemble determine the structure of these collaboration networks. We propose a model for the self-assembly of creative teams that has its basis in three parameters: team size, the fraction of newcomers in new productions, and the tendency of incumbents to repeat previous collaborations. The model suggests that the emergence of a large connected community of practitioners can be described as a phase transition. We find that team assembly mechanisms determine both the structure of the collaboration network and team performance for teams derived from both artistic and scientific fields.

Scientific articles are retracted at increasing rates, with the highest rates among top journals. Here we show that a single retraction triggers citation losses through an author's prior body of work. Compared to closely-matched control papers, citations fall by an average of 6.9% per year for each prior publication. These chain reactions are sustained on authors' papers (a) published up to a decade earlier and (b) connected within the authors' own citation network by up to 4 degrees of separation from the retracted publication. Importantly, however, citation losses among prior work disappear when authors self-report the error. Our analyses and results span the range of scientific disciplines.

Nodes in cooperative networks, such as those between plants and their pollinators or service providers and their contractors, form complex networks of interdependences. In these mutualistic networks, nodes that contribute to the nestedness of the network improve its stability. However, this study, using ecological data from 20 plant–pollinator networks and from socioeconomic networks, shows that these same nodes do not reap the benefits. In fact, the nodes that contribute the most to network persistence are also the most vulnerable to extinction. The architecture of mutualistic networks facilitates coexistence of individual participants by minimizing competition relative to facilitation1,2. However, it is not known whether this benefit is received by each participant node in proportion to its overall contribution to network persistence. This issue is critical to understanding the trade-offs faced by individual nodes in a network3,4,5. We address this question by applying a suite of structural and dynamic methods to an ensemble of flowering plant/insect pollinator networks. Here we report two main results. First, nodes contribute heterogeneously to the overall nested architecture of the network. From simulations, we confirm that the removal of a strong contributor tends to decrease overall network persistence more than the removal of a weak contributor. Second, strong contributors to collective persistence do not gain individual survival benefits but are in fact the nodes most vulnerable to extinction. We explore the generality of these results to other cooperative networks by analysing a 15-year time series of the interactions between designer and contractor firms in the New York City garment industry. As with the ecological networks, a firm's survival probability decreases as its individual nestedness contribution increases. Our results, therefore, introduce a new paradox into the study of the persistence of cooperative networks, and potentially address questions about the impact of invasive species in ecological systems and new competitors in economic systems.

Digital communication data has created opportunities to advance the knowledge of human dynamics in many areas, including national security, behavioral health, and consumerism. While digital data uniquely captures the totality of a person's communication, past research consistently shows that a subset of contacts makes up a person's “social network” of unique resource providers. To address this gap, we analyzed the correspondence between self-reported social network data and email communication data with the objective of identifying the dynamics in e-communication that correlate with a person's perception of a significant network tie. First, we examined the predictive utility of three popular methods to derive social network data from email data based on volume and reciprocity of bilateral email exchanges. Second, we observed differences in the response dynamics along self-reported ties, allowing us to introduce and test a new method that incorporates time-resolved exchange data. Using a range of robustness checks for measurement and misreporting errors in self-report and email data, we find that the methods have similar predictive utility. Although e-communication has lowered communication costs with large numbers of persons, and potentially extended our number of, and reach to contacts, our case results suggest that underlying behavioral patterns indicative of friendship or professional contacts continue to operate in a classical fashion in email interactions.

Successful animal systems often manage risk through synchronous behavior that spontaneously arises without leadership. In critical human systems facing risk, such as financial markets or military operations, our understanding of the benefits associated with synchronicity is nascent but promising. Building on previous work illuminating commonalities between ecological and human systems, we compare the activity patterns of individual financial traders with the simultaneous activity of other traders—an individual and spontaneous characteristic we call synchronous trading. Additionally, we examine the association of synchronous trading with individual performance and communication patterns. Analyzing empirical data on day traders’ second-to-second trading and instant messaging, we find that the higher the traders’ synchronous trading is, the less likely they are to lose money at the end of the day. We also find that the daily instant messaging patterns of traders are closely associated with their level of synchronous trading. This result suggests that synchronicity and vanguard technology may help traders cope with risky decisions in complex systems and may furnish unique prospects for achieving collective and individual goals.

Tracking the volume of keywords in Internet searches, message boards, or Tweets has provided an alternative for following or predicting associations between popular interest or disease incidences. Here, we extend that research by examining the role of e-communications among day traders and their collective understanding of the market. Our study introduces a general method that focuses on bundles of words that behave differently from daily communication routines, and uses original data covering the content of instant messages among all day traders at a trading firm over a 40-month period. Analyses show that two word bundles convey traders' understanding of same day market events and potential next day market events. We find that when market volatility is high, traders' communications are dominated by same day events, and when volatility is low, communications are dominated by next day events. We show that the stronger the traders' attention to either same day or next day events, the higher their collective trading performance. We conclude that e-communication among traders is a product of mass collaboration over diverse viewpoints that embodies unique information about their weak or strong understanding of the market.