Science of science

跳转到: 导航搜索


Science of science


Santo Fortunato1,2,, Carl T. Bergstrom3, Katy Börner2,4, James A. Evans5, Dirk Helbing6, Staša Milojević1, Alexander M. Petersen7, Filippo Radicchi1, Roberta Sinatra8,9,10, Brian Uzzi11,12, Alessandro Vespignani10,13,14, Ludo Waltman15, Dashun Wang11,12, Albert-László Barabási8,10,16,

Science 02 Mar Vol. 359, Issue 6379, eaao0185 DOI: 10.1126/science.aao0185

The whys and wherefores of SciSci

Scisci 的原因和原因

The science of science (SciSci) is based on a transdisciplinary approach that uses large data sets to study the mechanisms underlying the doing of science—from the choice of a research problem to career trajectories and progress within a field. In a Review, Fortunato et al. explain that the underlying rationale is that with a deeper understanding of the precursors of impactful science, it will be possible to develop systems and policies that improve each scientist's ability to succeed and enhance the prospects of science as a whole. 科学的科学(SciSci)是基于一种跨学科的方法, 它利用大量数据集来研究科学研究的潜在机制ーー从选择研究问题到职业发展轨迹和一个领域的进步。 在《评论》中, Fortunato 等人解释说, 其基本原理是, 随着对影响深远的科学前体有了更深刻的理解, 就有可能发展各种系统和政策, 提高每一位科学家成功的能力, 并增强整个科学的前景。

Structured Abstract



背景资料 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. 关于学术投入和产出的数字数据越来越多, 从研究资金、生产力和合作到论文引用和科学家的流动ーー提供了前所未有的机会来探索科学的结构和演变。 科学的科学(SciSci)提供了对不同地理和时间尺度的科学代理人之间相互作用的定量理解: 它提供了科学发现的创造性和起源的深刻见解, 最终目标是开发有可能加速科学发展的工具和政策。 在过去的十年里, SciSci 受益于自然科学、计算机科学和社会科学家的大量涌入, 他们共同开发了基于数据的大型能力, 用于进行实证分析和生成模型, 捕捉科学、机构和劳动力的展开。 Scisci 的价值主张在于, 通过对推动科学成功的因素有了更深入的理解, 我们就能更有效地解决环境、社会和技术问题。


进步 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. 科学可以被描述为一个复杂的、自我组织的、不断发展的学者、项目、论文和思想网络。 通过对协作网络的研究, 以及通过对引文网络的研究, 揭示了新科学领域的出现模式。 微观模型已经追踪到引文积累的动态, 使我们能够预测单个论文的未来影响。 Scisci 揭示了科学家在推进自己的事业和科学前景时所面临的选择和权衡。 例如, 测量结果表明, 学者们不愿承担风险, 他们更愿意研究与他们目前的专业知识相关的主题, 这限制了未来发现的潜力。 那些愿意打破这种模式的人从事风险更高的职业, 但是他们更有可能取得重大突破。 总的来说, 影响最大的科学是基于传统的组合的先前的工作, 但特点不寻常的组合。 最后, 随着研究中心转移到团队中, SciSci 越来越关注团队研究的影响, 发现小团队倾向于利用老的和不那么普遍的思想来扰乱科学和技术。 相比之下, 大团队倾向于发展最近流行的想法, 获得高, 但往往是短暂的影响。


展望 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. 为科学家、机构和思想之间的关系结构提供了深入的定量理解, 因为它有助于确定科学发现的基本机制。 这些多学科数据驱动的努力补充了科学的科学和经济学和社会学等相关领域的贡献。 虽然 SciSci 寻求长期存在的适用于各种科学领域的普遍法律和机制, 但今后的一项基本挑战是, 如何解释不同领域和国家之间不可否认的文化、习惯和偏好的差异。 这种变化使得一些跨领域的见解难以理解, 相关的科学政策难以实施。 每个学科所特有的问题、数据和技能之间的差异表明, 可以从针对特定领域的 SciSci 研究中获得进一步的见解, 这些研究模拟和确定适合个别研究领域需要的机会。


摘要 Identifying fundamental drivers of science and developing predictive models to capture its evolution are instrumental for the design of policies that can improve the scientific enterprise—for example, through enhanced career paths for scientists, better performance evaluation for organizations hosting research, discovery of novel effective funding vehicles, and even identification of promising regions along the scientific frontier. The science of science uses large-scale data on the production of science to search for universal and domain-specific patterns. Here, we review recent developments in this transdisciplinary field. 确定科学的基本驱动因素并开发预测模型以捕捉其演变, 有助于制定能够改进科学事业的政策, 例如, 通过加强科学家的职业道路、对主办研究的组织进行更好的业绩评价、发现新的有效供资工具, 甚至查明科学前沿的有希望的区域。 科学科学利用大规模的科学生产数据来寻找普遍和特定领域的模式。 在这里, 我们回顾了这个跨学科领域的最新发展。

The deluge of digital data on scholarly output offers unprecedented opportunities to explore patterns characterizing the structure and evolution of science. The science of science (SciSci) places the practice of science itself under the microscope, leading to a quantitative understanding of the genesis of scientific discovery, creativity, and practice and developing tools and policies aimed at accelerating scientific progress. 关于学术产出的数字数据的泛滥为探索科学结构和进化的模式提供了前所未有的机会。 科学的科学(SciSci)将科学实践本身置于显微镜下, 从而对科学发现、创造力和实践的起源有了定量的认识, 并制定了旨在加速科学进步的工具和政策。

The emergence of SciSci has been driven by two key factors. The first is data availability. In addition to the proprietary Web of Science (WoS), the historic first citation index (1), multiple data sources are available today (Scopus, PubMed, Google Scholar, Microsoft Academic, the U.S. Patent and Trademark Office, and others). Some of these sources are freely accessible, covering millions of data points pertaining to scientists and their output and capturing research from all over the world and all branches of science. Second, SciSci has benefited from an influx of and collaborations among natural, computational, and social scientists who have developed big data–based capabilities and enabled critical tests of generative models that aim to capture the unfolding of science, its institutions, and its workforce. Scisci 的出现受到两个关键因素的驱使。 首先是数据的可用性。 除了专有的科学网络(WoS) , 历史上第一个引用索引(1) , 今天还有多个数据源(Scopus, PubMed, Google Scholar, Microsoft Academic, the u.s. Patent and Trademark Office, and others)。 其中一些资料来源可以自由获取, 涵盖与科学家及其产出有关的数百万个数据点, 并记录了来自世界各地和所有科学分支的研究成果。 其次, SciSci 受益于自然科学家、计算机科学家和社会科学家的大量涌入和合作, 他们开发了基于数据的大型能力, 并对生成模型进行了关键的测试, 目的是抓住科学、其机构和劳动力的展开。

One distinctive characteristic of this emerging field is how it breaks down disciplinary boundaries. SciSci integrates findings and theories from multiple disciplines and uses a wide range of data and methods. From scientometrics, it takes the idea of measuring science from large-scale data sources; from the sociology of science, it adopts theoretical concepts and social processes; and from innovation studies, it explores and identifies pathways through which science contributes to invention and economic change. SciSci relies on a broad collection of quantitative methods, from descriptive statistics and data visualization to advanced econometric methods, network science approaches, machine-learning algorithms, mathematical analysis, and computer simulation, including agent-based modeling. The value proposition of SciSci hinges on the hypothesis that with a deeper understanding of the factors behind successful science, we can enhance the prospects of science as a whole to more effectively address societal problems. 这个新兴领域的一个显著特点就是如何打破学科界限。 Scisci 集成了多个学科的发现和理论, 并使用了广泛的数据和方法。 从科学科学的角度来看, 它采用了从大规模数据源测量科学的想法; 从科学社会学, 它采用理论概念和社会过程; 从创新研究中, 它探索并确定了科学促进发明和经济变革的途径。 Scisci 依靠大量的定量方法, 从描述统计学和数据可视化到先进的计量经济学方法、网络科学方法、机器学习算法、数学分析和计算机模拟, 包括基于代理的建模。 Scisci 的价值主张取决于这样一个假设, 即随着我们对成功科学背后的因素有了更深入的理解, 我们可以提高整个科学的前景, 从而更有效地解决社会问题。

Networks of scientists, institutions, and ideas 科学家、机构和思想网络

Contemporary science is a dynamical system of undertakings driven by complex interactions among social structures, knowledge representations, and the natural world. Scientific knowledge is constituted by concepts and relations embodied in research papers, books, patents, software, and other scholarly artifacts, organized into scientific disciplines and broader fields. These social, conceptual, and material elements are connected through formal and informal flows of information, ideas, research practices, tools, and samples. Science can thus be described as a complex, self-organizing, and constantly evolving multiscale network. 当代科学是由社会结构、知识表征和自然世界之间复杂的相互作用所驱动的事业的动力系统。 科学知识是由研究论文、书籍、专利、软件和其他学术文献所体现的概念和关系构成的, 它们被组织成科学分支和更广阔的领域。 这些社会性、概念性和物质性的元素通过正式和非正式的信息、思想、研究实践、工具和样本的流动。 因此, 科学可以被描述为一个复杂的、自我组织的、不断演变的多尺度网络。

Early studies discovered an exponential growth in the volume of scientific literature (2), a trend that continues with an average doubling period of 15 years (Fig. 1). Yet, it would be naïve to equate the growth of the scientific literature with the growth of scientific ideas. Changes in the publishing world, both technological and economic, have led to increasing efficiency in the production of publications. Moreover, new publications in science tend to cluster in discrete areas of knowledge (3). Large-scale text analysis, using phrases extracted from titles and abstracts to measure the cognitive extent of the scientific literature, have found that the conceptual territory of science expands linearly with time. In other words, whereas the number of publications grows exponentially, the space of ideas expands only linearly (Fig. 1) (4). 早期的研究发现科学文献的数量有一个指数增长(2) , 这种趋势持续下去, 平均增加了15年(图1)。 然而, 将科学文献的增长等同于科学思想的发展。 出版界的变化, 包括技术和经济方面的变化, 都提高了出版物的制作效率。 此外, 科学领域的新出版物往往集中在不同的知识领域(3)。 大规模的文本分析, 利用从标题和摘要中提取的短语来衡量科学文献的认知程度, 发现科学的概念领域随着时间的推移呈线性增长。 换句话说, 尽管出版物的数量呈指数增长, 思想空间只是线性地扩展(图1)(4)。

Frequently occurring words and phrases in article titles and abstracts propagate via citation networks, punctuated by bursts corresponding to the emergence of new paradigms (5). By applying network science methods to citation networks, researchers are able to identify communities as defined by subsets of publications that frequently cite one another (6). These communities often correspond to groups of authors holding a common position regarding specific issues (7) or working on the same specialized subtopics (8). Recent work focusing on biomedical science has illustrated how the growth of the literature reinforces these communities (9). As new papers are published, associations (hyperedges) between scientists, chemicals, diseases, and methods (“things,” which are the nodes of the network) are added. Most new links fall between things only one or two steps away from each other, implying that when scientists choose new topics, they prefer things directly related to their current expertise or that of their collaborators. This densification suggests that the existing structure of science may constrain what will be studied in the future. 文章标题中频繁出现的单词和短语以及通过引文网络传播的抽象词汇, 以及与新范式的出现相对应的爆发。 通过将网络科学方法应用到引文网络中, 研究人员能够识别出经常互相引用的出版物子集所定义的群落(6)。 这些社区往往与对具体问题持共同立场的作者群体相对应, 或者在同样的专门分专题(8)上工作。 近期关注生物医学科学的工作表明, 文献的增长如何加强这些社区(9)。 随着新论文的发表, 科学家、化学物质、疾病和方法之间的关联(超边缘)("事物"是网络的节点)。 大多数新的链接都是在相距只有一两步的事情之间, 这意味着当科学家选择新的主题时, 他们更喜欢与他们当前的专业知识或者他们的合作者的专业知识直接相关的东西。 这种反思表明, 现有的科学结构可能会限制未来的研究。

Densification at the boundaries of science is also a signal of transdisciplinary exploration, fusion, and innovation. A life-cycle analysis of eight fields (10) shows that successful fields undergo a process of knowledge and social unification that leads to a giant connected component in the collaboration network, corresponding to a sizeable group of regular coauthors. A model in which scientists choose their collaborators through random walks on the coauthorship network successfully reproduces author productivity, the number of authors per discipline, and the interdisciplinarity of papers and authors (11). 科学界限上的密集化也是跨学科探索、融合和创新的信号。 一个由8个领域组成的生命週期评估(10)表明, 成功的领域经历了一个知识和社会统一的过程, 这将导致协作网络中一个巨大的连接元件(图论) , 相当于一个相当大的合作者群体。 科学家通过在合作者网络上的随机行走来选择合作者的模型成功地重现了作者的生产力, 每个学科的作者数量, 以及论文和作者的科际整合(11)。

Problem selection

问题选择 How do scientists decide which research problems to work on? Sociologists of science have long hypothesized that these choices are shaped by an ongoing tension between productive tradition and risky innovation (12, 13). Scientists who adhere to a research tradition in their domain often appear productive by publishing a steady stream of contributions that advance a focused research agenda. But a focused agenda may limit a researcher’s ability to sense and seize opportunities for staking out new ideas that are required to grow the field’s knowledge. For example, a case study focusing on biomedical scientists choosing novel chemicals and chemical relationships shows that as fields mature, researchers tend to focus increasingly on established knowledge (3). Although an innovative publication tends to result in higher impact than a conservative one, high-risk innovation strategies are rare, because the additional reward does not compensate for the risk of failure to publish at all. Scientific awards and accolades appear to function as primary incentives to resist conservative tendencies and encourage betting on exploration and surprise (3). Despite the many factors shaping what scientists work on next, macroscopic patterns that govern changes in research interests along scientific careers are highly reproducible, documenting a high degree of regularity underlying scientific research and individual careers (14). 科学家如何决定哪些研究问题需要解决? 科学的社会学家长期以来一直假设这些选择是由于生产传统和风险创新之间持续的紧张关系(12,13)。 在他们的领域坚持研究传统的科学家通常会出版稳定的贡献流, 推进一个重点突出的研究议程, 从而显得富有成效。 但是一个专注的议程可能会限制研究人员感知和抓住机会的能力, 因为这些新思想是培养该领域知识所必需的。 例如, 一个以生物医学科学家选择新的化学物质和化学关系的案例研究表明, 随着领域的成熟, 研究人员倾向于越来越多地关注已有的知识(3)。 虽然创新性出版物往往比保守的出版物产生更大的影响, 但高风险的创新战略很少, 因为额外的奖励并不能弥补完全不发表的风险。 科学奖项和荣誉似乎是抵制保守倾向的主要动机, 鼓励人们在探索和惊喜上下赌注(3)。 尽管影响科学家下一步工作方向的因素很多, 但是在科学职业中支配研究兴趣变化的宏观模式是高度可复制的, 记录了科学研究和个人职业的高度规律性(14)。

Scientists’ choice of research problems affects primarily their individual careers and the careers of those reliant on them. Scientists’ collective choices, however, determine the direction of scientific discovery more broadly (Fig. 2). Conservative strategies (15) serve individual careers well but are less effective for science as a whole. Such strategies are amplified by the file drawer problem (16): Negative results, at odds with established hypotheses, are rarely published, leading to a systemic bias in published research and the canonization of weak and sometimes false facts (17). More risky hypotheses may have been tested by generations of scientists, but only those successful enough to result in publications are known to us. One way to alleviate this conservative trap is to urge funding agencies to proactively sponsor risky projects that test truly unexplored hypotheses and take on special interest groups advocating for particular diseases. Measurements show that the allocation of biomedical resources in the United States is more strongly correlated to previous allocations and research than to the actual burden of diseases (18), highlighting a systemic misalignment between biomedical needs and resources. This misalignment casts doubts on the degree to which funding agencies, often run by scientists embedded in established paradigms, are likely to influence the evolution of science without introducing additional oversight, incentives, and feedback. 科学家对研究问题的选择主要影响到他们的个人事业和依赖他们的人的职业生涯。 然而, 科学家的集体选择更广泛地决定了科学发现的方向(图2)。 保守策略(15)很好地服务于个人事业, 但对整个科学而言效果不佳。 这些策略被文件抽屉问题(16)放大: 负面的结果, 与既定的假设不一致, 很少发表, 导致出版研究中的系统性偏见和虚假事实的封建(17)。 更有风险的假设可能已经被几代科学家测试过了, 但是只有那些成功的人才知道出版物。 缓解这种保守陷阱的一个办法是, 敦促供资机构积极主动地赞助那些真正检验未经探索的假设的风险项目, 并接纳那些主张特定疾病的特殊利益集团。 测量结果表明, 美国生物医学资源的分配与以前的分配和研究之间的关系更为密切, 而不是与疾病的实际负担相关(18) , 这突出表明生物医学需求和资源之间存在系统性的不协调。 这种失调使人们怀疑经常由嵌入在既定范式中的科学家管理的供资机构在多大程度上可能影响科学的发展, 而不会引入额外的监督、激励和反馈。


新奇 Analyses of publications and patents consistently reveal that rare combinations in scientific discoveries and inventions tend to garner higher citation rates (3). Interdisciplinary research is an emblematic recombinant process (19); hence, the successful combination of previously disconnected ideas and resources that is fundamental to interdisciplinary research often violates expectations and leads to novel ideas with high impact (20). Nevertheless, evidence from grant applications shows that, when faced with new ideas, expert evaluators systematically give lower scores to truly novel (21–23) or interdisciplinary (24) research proposals. 对出版物和专利的分析一致表明, 科学发现和发明中罕见的组合往往获得更高的引用率(3)。 科际整合是一个具有象征意义的重组过程(19) ; 因此, 先前与科际整合无关的思想和资源成功地结合在一起, 往往违背了人们的期望, 并导致了具有高度影响力的新想法(20)。 不过, 来自赠款申请的证据表明, 在面对新的想法时, 专家评价者系统地给出真正新颖的(21-23)或跨学科(24)研究提案的分数较低。 The highest-impact science is primarily grounded in conventional combinations of prior work, yet it simultaneously features unusual combinations (25–27). Papers of this type are twice as likely to receive high citations (26). In other words, a balanced mixture of new and established elements is the safest path toward successful reception of scientific advances. 影响最大的科学主要是基于以往工作的传统组合, 但它同时具有不同寻常的组合(25-27)。 这种类型的论文获得高引文的可能性是其他论文的两倍(26)。 换句话说, 新的和既定的元素的平衡混合是成功接受科学进步的最安全的途径。

Career dynamics

职业动态 Individual academic careers unfold in the context of a vast market for knowledge production and consumption (28). Consequently, scientific careers have been examined not only in terms of individual incentives and marginal productivity (i.e., relative gain versus effort) (29), but also institutional incentives (30, 31) and competition (32). This requires combining large repositories of high-resolution individual, geographic, and temporal metadata (33) to construct representations of career trajectories that can be analyzed from different perspectives. For example, one study finds that funding schemes that are tolerant of early failure, which reward long-term success, are more likely to generate high-impact publications than grants subject to short review cycles (31). Interacting systems with competing time scales are a classic problem in complex systems science. The multifaceted nature of science is motivation for generative models that highlight unintended consequences of policies. For example, models of career growth show that nontenure (short-term) contracts are responsible for productivity fluctuations, which often result in a sudden career death (29). 个人学术生涯是在知识生产和消费的巨大市场中展开的(28)。 因此, 科学职业不仅从个人激励和边际生产力(即相对增益与努力)(29)来审视科学职业, 而且还审查了体制激励(30,31)和竞争(32)。 这就需要将大量高分辨率的个人、地理和时间元数据(33)组合起来, 构建可以从不同角度分析的职业轨迹的表述。 例如, 一项研究发现, 容忍早期失败的供资计划比审查周期短的赠款更有可能产生影响较大的出版物(31)。 具有相互竞争时间尺度的互动系统是复杂系统科学中的一个经典问题。 科学的多面性是生成模型的动力, 这些模型突出政策的意外后果。 例如, 职业发展模式表明, 非任期(短期)合同是造成生产力波动的原因, 这往往导致职业突然死亡(29)。

Gender inequality in science remains prevalent and problematic (34). Women have fewer publications (35–37) and collaborators (38) and less funding (39), and they are penalized in hiring decisions when compared with equally qualified men (40). The causes of these gaps are still unclear. Intrinsic differences in productivity rates and career length can explain the differences in collaboration patterns (38) and hiring rates (35) between male and female scientists. On the other hand, experimental evidence shows that biases against women occur at very early career stages. When gender was randomly assigned among the curricula vitae of a pool of applicants, the hiring committee systematically penalized female candidates (40). Most studies so far have focused on relatively small samples. Improvements in compiling large-scale data sets on scientific careers, which leverage information from different sources (e.g., publication records, grant applications, and awards), will help us gain deeper insight into the causes of inequality and motivate models that can inform policy solutions. 科学领域的性别不平等仍然普遍存在, 问题重重(34)。 妇女的出版物(35-37)和合作者(38)较少, 资金较少(39) , 与同等资格的男子相比, 她们在雇用决定方面受到惩罚(40)。 造成这些差距的原因尚不清楚。 生产率和职业生涯长度的内在差异可以解释男女科学家之间在协作模式(38)和雇佣率(35)方面的差异。 另一方面, 实验证据表明, 对妇女的偏见发生在非常早期的职业阶段。 如果将性别问题随机分配给一批申请者的履历时, 招聘委员会就会系统地惩罚女性候选人(40名)。 到目前为止, 大多数研究都集中在相对较小的样本上。 改进编制关于科学职业的大规模数据集, 利用来自不同来源的信息(例如出版记录、赠款申请和奖励) , 将有助于我们更深入地了解不平等的原因, 并激发能够为政策解决方案提供信息的模式。

Scientists’ mobility is another important factor offering diverse career opportunities. Most mobility studies have focused on quantifying the brain drain and gain of a country or a region (41, 42), especially after policy changes. Research on individual mobility and its career effect remains scant, however, primarily owing to the difficulty of obtaining longitudinal information about the movements of many scientists and accounts of the reasons underlying mobility decisions. Scientists who left their country of origin outperformed scientists who did not relocate, according to their citation scores, which may be rooted in a selection bias that offers better career opportunities to better scientists (43, 44). Moreover, scientists tend to move between institutions of similar prestige (45). Nevertheless, when examining changes in impact associated with each move as quantified by citations, no systematic increase or decrease was found, not even when scientists moved to an institution of considerably higher or lower rank (46). In other words, it is not the institution that creates the impact; it is the individual researchers that make an institution. 科学家的流动性是提供多样职业机会的另一个重要因素。 大多数流动性研究的重点是量化一个国家或地区的人才流失和获益(41,42) , 特别是在政策改变之后。 但是, 关于个人流动性及其职业影响的研究仍然很少, 主要原因是很难获得关于许多科学家行动的纵向信息, 以及对作出流动决定的原因作出说明。 根据科学家们的引文评分, 那些离开起源国的科学家的表现胜过那些没有搬迁的科学家, 而这些评分可能源于一种选择偏见, 这种偏见为更优秀的科学家提供了更好的职业机会(43,44)。 此外, 科学家倾向于在声望相似的机构之间迁移(45)。 尽管如此, 在研究引用量化的与每一行动相关的影响变化时, 也没有发现系统性的增加或减少, 甚至在科学家迁往一个级别相当高或较低的机构(46)时也没有发现。 换句话说, 造成这种影响的并不是机构本身, 而是由个体研究人员组成的机构。

Another potentially important career factor is reputation—and the dilemma that it poses for manuscript review, proposal evaluation, and promotion decisions. The reputation of paper authors, measured by the total citations of their previous output, markedly boosts the number of citations collected by that paper in the first years after publication (47). After this initial phase, however, impact depends on the reception of the work by the scientific community. This finding, along with the work reported in (46), suggests that, for productive scientific careers, reputation is less of a critical driver for success than talent, hard work, and relevance. 另一个潜在的重要的职业因素是声誉——以及它给手稿评审、求职评估和晋升决定带来的困境。 论文作者的声誉, 以他们以前产出的总引文来衡量, 显著提高了该论文在出版后第一年收集的引文数量(47)。 然而, 在这个初始阶段之后, 影响取决于科学界对这项工作的接受程度。 这一发现, 连同(46)中所报道的工作一起, 表明, 对于高效的科学职业来说, 声誉并不是成功的关键驱动因素, 而是人才、努力工作和相关性。

A policy-relevant question is whether creativity and innovation depend on age or career stage. Decades of research on outstanding researchers and innovators concluded that major breakthroughs take place relatively early in a career, with a median age of 35 (48). In contrast, recent work shows that this well-documented propensity of early-career discoveries is fully explained by productivity, which is high in the early stages of a scientist’s career and drops later (49). In other words, there are no age patterns in innovation: A scholar’s most cited paper can be any of his or her papers, independently of the age or career stage when it is published (Fig. 3). A stochastic model of impact evolution also indicates that breakthroughs result from a combination of the ability of a scientist and the luck of picking a problem with high potential (49). 一个与政策相关的问题是, 创造力和创新是否取决于年龄或职业阶段。 几十年来对杰出研究人员和创新者的研究得出结论, 在职业生涯中, 主要的突破性进展相对较早, 中位年龄为35岁(48岁)。 相比之下, 最近的研究表明, 早期职业发现的这种有据可查的倾向完全可以用生产力来解释, 生产力在科学家职业生涯的早期阶段很高, 后来下降(49)。 换句话说, 创新中没有年龄模式: 学者引用最多的论文可以是他或她的任何论文, 不论其年龄或职业阶段何时发表(图3)。 影响演化的随机模型也表明, 科学家的能力与挑选高潜力问题的运气(49)相结合的结果。

Team science

团队科学 During past decades, reliance on teamwork has increased, representing a fundamental shift in the way that science is done. A study of the authorship of 19.9 million research articles and 2.1 million patents reveals a nearly universal shift toward teams in all branches of science (50) (Fig. 4). For example, in 1955, science and engineering teams authored about the same number of papers as single authors. Yet by 2013, the fraction of team-authored papers increased to 90% (51). 在过去的几十年里, 对团队合作的依赖增加了, 这代表着科学发展方式的根本转变。 一项关于1990万篇研究论文和210万项专利的研究显示, 几乎所有科学分支的团队都发生了普遍的转变(图4)。 例如, 1955年, 科学和工程团队作为单一作者撰写了同样数量的论文。 然而, 到2013年, 团队撰写论文的比例增加到90% (51)。

Nowadays, a team-authored paper in science and engineering is 6.3 times more likely to receive 1000 citations or more than a solo-authored paper, a difference that cannot be explained by self-citations (50, 52). One possible reason is a team's ability to come up with more novel combinations of ideas (26) or to produce resources that are later used by others (e.g., genomics). Measurements show that teams are 38% more likely than solo authors to insert novel combinations into familiar knowledge domains, supporting the premise that teams can bring together different specialties, effectively combining knowledge to prompt scientific breakthroughs. Having more collaborations means greater visibility through a larger number of coauthors, who will likely introduce the work to their networks, an enhanced impact that may partially compensate for the fact that credit within a team must be shared with many colleagues (29). 如今, 一个科学和工程领域的研究小组撰写的论文获得1000次引文或超过一篇独立论文的可能性高出6.3倍, 这种差异不能用自我引用来解释(50,52)。 一个可能的原因是, 一个团队有能力想出更多新颖的想法组合(26) , 或者生产其他人后来使用的资源(例如, 基因组学)。 测量结果表明, 团队比独立作者更有可能将新颖的组合插入熟悉的知识领域, 支持这样的前提, 即团队可以把不同的专业结合起来, 有效地结合知识来促进科学突破。 拥有更多的合作意味着通过更多的合作者来提高能见度, 他们可能会把这项工作介绍给他们的网络, 这样的影响可以部分弥补一个事实, 即团队内部的信贷必须与许多同事分享(29)。

Work from large teams garners, on average, more citations across a wide variety of domains. Research suggests that small teams tend to disrupt science and technology with new ideas and opportunities, whereas large teams develop existing ones (53). Thus, it may be important to fund and foster teams of all sizes to temper the bureaucratization of science (28). 平均来说, 大型团队的工作在很多领域都有更多的引用。 研究表明, 小团队倾向于用新的想法和机会扰乱科学和技术, 而大型团队则发展现有的(53个)。 因此, 为各种规模的团队提供资金和培训, 以缓和科学的官僚化(28)。

Teams are growing in size, increasing by an average of 17% per decade (50, 54), a trend underlying a fundamental change in team compositions. Scientific teams include both small, stable “core” teams and large, dynamically changing extended teams (55). The increasing team size in most fields is driven by faster expansion of extended teams, which begin as small core teams but subsequently attract new members through a process of cumulative advantage anchored by productivity. Size is a crucial determinant of team survival strategies: Small teams survive longer if they maintain a stable core, but larger teams persist longer if they manifest a mechanism for membership turnover (56). 团队规模不断扩大, 平均每十年增长17% (50,54) , 这是团队组成发生根本变化的潜在趋势。 科学团队包括小型、稳定的"核心"团队和大型、动态变化的扩展团队(55个)。 在大多数领域, 团队规模的增加是由扩展团队的快速扩展所驱动的, 这些团队一开始是小型核心团队, 但后来通过以生产力为基础的累积优势进程吸引新成员。 规模是团队生存策略的一个关键决定因素: 如果小型团队保持稳定的核心, 他们的生存时间会更长, 但是如果更大的团队表现出成员流动的机制, 他们会坚持更长的时间(56)。

As science has accelerated and grown increasingly complex, the instruments required to expand the frontier of knowledge have increased in scale and precision. The tools of the trade become unaffordable to most individual investigators, but also to most institutions. Collaboration has been a critical solution, pooling resources to scientific advantage. The Large Hadron Collider at CERN, the world’s largest and most powerful particle collider, would have been unthinkable without collaboration, requiring more than 10,000 scientists and engineers from more than 100 countries. There is, however, a trade-off with increasing size that affects the value and risk associated with “big science” (2). Although it may be possible to solve larger problems, the burden of reproducibility may require duplicating initial efforts, which may not be practically or economically feasible. 随着科学的加速发展和日益复杂, 扩大知识领域所需的工具的规模和精确度都有所提高。 对于大多数个体调查人员来说, 这种交易的工具已经变得无法承受, 而且对大多数机构来说也是如此。 合作一直是一个关键的解决方案, 把资源集中到科学优势上。 如果没有合作, 欧洲核子研究中心这个世界上最大、最强大的粒子对撞机, 如果没有合作的话, 将是不可想象的, 这需要来自100多个国家的1万多名科学家和工程师。 然而, 随着规模的扩大, 存在一种权衡, 它影响到与"大科学"(2)相关的价值和风险。 虽然有可能解决较大的问题, 但重复性的负担可能需要重复最初的努力, 这种努力在实际上或经济上可能不可行。

Collaborators can have a large effect on scientific careers. According to recent studies (57, 58), scientists who lose their star collaborators experience a substantial drop in their productivity, especially if the lost collaborator was a regular coauthor. Publications involving extremely strong collaborators gain 17% more citations on average, pointing to the value of career partnership (59). 合作者可以对科学事业产生巨大影响。 根据最近的研究(57,58) , 失去明星合作者的科学家的生产力会大幅下降, 特别是如果失败的合作者是一个普通的合作者。 涉及极其强大合作者的出版物平均多获得17% 的引文, 这表明了职业伙伴关系的价值(59)。

Given the increasing number of authors on the average research paper, who should and does gain the most credit? The canonical theory of credit (mis)allocation in science is the Matthew effect (60), in which scientists of higher statuses involved in joint work receive outsized credit for their contributions. Properly allocating individual credit for a collaborative work is difficult because we cannot easily distinguish individual contributions (61). It is possible, however, to inspect the cocitation patterns of the coauthors’ publications to determine the fraction of credit that the community assigns to each coauthor in a publication (62). 鉴于平均研究论文的作者越来越多, 谁应该也确实获得了最多的赞誉? 科学中的信用分配规范理论是马太效应(60) , 其中参与联合工作的高级别科学家的贡献获得了巨额的信贷。 为协作工作适当分配个人信贷是困难的, 因为我们不能轻易区分个人贡献(61)。 然而, 可以检查共同作者出版物的背诵模式, 以确定社区在出版物中向每个共同作者分配的信用分数(62)。

Citation dynamics

引用动力学 Scholarly citation remains the dominant measurable unit of credit in science. Given the reliance of most impact metrics on citations (63–66), the dynamics of citation accumulation have been scrutinized by generations of scholars. From foundational work by Price (67), we know that the distribution of citations for scientific papers is highly skewed: Many papers are never cited, but seminal papers can accumulate 10,000 or more citations. This uneven citation distribution is a robust, emergent property of the dynamics of science, and it holds when papers are grouped by institution (68). If the number of citations of a paper is divided by the average number of citations collected by papers in the same discipline and year, the distribution of the resulting score is essentially indistinguishable for all disciplines (69, 70) (Fig. 5A). This means that we can compare the impact of papers published in different disciplines by looking at their relative citation values. For example, a paper in mathematics collecting 100 citations represents a higher disciplinary impact than a paper in microbiology with 300 citations. 在科学领域, 学术引文仍然是可衡量的主要信用单位。 由于大多数影响指标都依赖于引文(63-66) , 引文积累的动态已经被几代学者仔细研究过了。 从普莱斯的基础著作中, 我们知道科学论文引文的分布是非常不平衡的: 许多论文从来没有被引用过, 但是开创性的论文可以累积10,000或更多的引文。 这种不均匀的引文分布是科学动力学的一个强大的突现性质, 当文件被机构分组时, 它就会保持不变。 如果一份文件的引文次数除以同一学科和年份文件收集的平均引文次数, 则所得分数的分布对所有学科基本上是无法区分的(69,70)(图5A)。 这意味着我们可以通过观察在不同学科发表的论文的相对引用值来比较其影响。 例如, 一篇收集100次引文的数学论文对学科的影响比微生物学中的一篇论文的300次引文有更高的学科影响。

The tail of the citation distribution, capturing the number of high-impact papers, sheds light on the mechanisms that drive the accumulation of citations. Recent analyses show that it follows a power law (71–73). Power-law tails can be generated through a cumulative advantage process (74), known as preferential attachment in network science (75), suggesting that the probability of citing a paper grows with the number of citations that it has already collected. Such a model can be augmented with other characteristic features of citation dynamics, such as the obsolescence of knowledge, decreasing the citation probability with the age of the paper (76–79), and a fitness parameter, unique to each paper, capturing the appeal of the work to the scientific community (77, 78). Only a tiny fraction of papers deviate from the pattern described by such a model—some of which are called “sleeping beauties,” because they receive very little notice for decades after publication and then suddenly receive a burst of attention and citations (80, 81). 引文分发的尾巴记录了高影响力论文的数量, 揭示了引起引文积累的机制。 最近的分析表明, 它遵循一项权力法(71-73)。 电力法尾巴可以通过累积优势过程(74)产生, 这个过程被称为网络科学中的优先附件(75) , 这表明引用纸张的可能性随着已经收集到的引用次数而增加。 这样的模型可以通过引文动力学的其他特征增强, 例如知识的淘汰, 随着论文的年龄(76-79)降低了引文概率, 以及一个适合每篇论文的参数, 捕捉了这项工作对科学界的吸引力(77,78)。 只有很小一部分论文偏离了这种模式所描述的模式, 其中有些被称为"睡美人", 因为在出版几十年后, 它们几乎没有收到什么通知, 然后突然得到一阵轰动的注意和引用(80,81)。

The generative mechanisms described above can be used to predict the citation dynamics of individual papers. One predictive model (77) assumes that the citation probability of a paper depends on the number of previous citations, an obsolescence factor, and a fitness parameter (Fig. 5, B and C). For a given paper, one can estimate the three model parameters by fitting the model to the initial portion of the citation history of the paper. The long-term impact of the work can be extrapolated (77). Other studies have identified predictors of the citation impact of individual papers (82), such as journal impact factor (72). It has been suggested that the future h-index (83) of a scientist can be accurately predicted (84), although the predictive power is reduced when accounting for the scientist’s career stage and the cumulative, nondecreasing nature of the h-index (85). Eliminating inconsistencies in the use of quantitative evaluation metrics in science is crucial and highlights the importance of understanding the generating mechanisms behind commonly used statistics. 上述生成机制可以用来预测单个论文的引文动态。 一个预测模型(77)假设论文的引用概率取决于以前引用的次数、过时因子和适应参数(图5、 b 和 c)。 对于给定的论文, 我们可以通过将模型与论文引用历史的最初部分相匹配来估计三个模型参数。 这项工作的长期影响可以外推(77)。 其他研究已经确定了单个论文引用影响的预测因子(82) , 如期刊影响因子(72)。 虽然在考虑科学家的职业生涯阶段以及 h 指数(85)的累积性、非递减性, 但可以准确地预测一位科学家的未来 h 指数(83)(84)。 消除在科学中使用定量评价指标方面的不一致, 这是至关重要的, 并强调了了解通常使用的统计数据背后的生成机制的重要性。


展望 Despite the discovery of universals across science, substantial disciplinary differences in culture, habits, and preferences make some cross-domain insights difficult to appreciate within particular fields and associated policies challenging to implement. The differences among the questions, data, and skills required by each discipline suggest that we may gain further insights from domain-specific SciSci studies that model and predict opportunities adapted to the needs of each field. For young scientists, the results of SciSci offer actionable insights about past patterns, helping guide future inquiry within their disciplines (Box 1). 尽管在科学上发现了普遍性, 但是在文化、习惯和偏好上存在巨大的学科差异, 使得某些跨领域的见解难以在特定领域中得到理解, 而且相关政策难以实施。 每个学科所要求的问题、数据和技能之间的差异表明, 我们可以从针对特定领域的 SciSci 研究中获得进一步的见解, 这些研究模型和预测适合每个领域需要的机会。 对于年轻科学家来说, SciSci 的研究结果提供了关于过去模式的可行的见解, 帮助指导未来的研究在他们的学科(框1)。 Lessons from SciSci.

来自 SciSci 的教训

Innovation and tradition: Left bare, truly innovative and highly interdisciplinary ideas may not reach maximum scientific impact. To enhance their impact, novel ideas should be placed in the context of established knowledge (26). 创新与传统: 赤裸裸的、真正的创新和高度跨学科的想法可能不会达到最大的科学影响。 为了提高其影响力, 应在已经确定的知识范围内提出新的想法(26)。 Persistence: A scientist is never too old to make a major discovery, as long as he or she stays productive (49). 坚持不懈: 一个科学家只要保持生产力, 就永远不会太老, 不会有重大发现。 Collaboration: Research is shifting to teams, so engaging in collaboration is beneficial. Works by small teams tend to be more disruptive, whereas those by big teams tend to have more impact (4, 50, 53). 合作: 研究正在转向团队, 所以合作是有益的。 小团队的工作往往更具破坏性, 而大团队的工作往往有更大的影响(4,50,53)。 Credit: Most credit will go to the coauthors with the most consistent track record in the domain of the publication (62). 图片来源: 大多数信用都将归功于出版物领域中记录最一致的共同作者(62)。 Funding: Although review panels acknowledge innovation, they ultimately tend to discount it. Funding agencies should ask reviewers to assess innovation, not only expected success (24). 资金: 虽然评审小组承认创新, 但他们最终倾向于忽略创新。 资助机构应该要求评审人员评估创新, 而不仅仅是预期的成功(24)。

The contribution of SciSci is a detailed understanding of the relational structure between scientists, institutions, and ideas, a crucial starting point that facilitates the identification of fundamental generating processes. Together, these data-driven efforts complement contributions from related research domains such as the economics (30) and sociology of science (60, 86). Causal estimation is a prime example, in which econometric matching techniques demand and leverage comprehensive data sources in the effort to simulate counterfactual scenarios (31, 42). Assessing causality is one of the most needed future developments in SciSci: Many descriptive studies reveal strong associations between structure and outcomes, but the extent to which a specific structure “causes” an outcome remains unexplored. Engaging in tighter partnerships with experimentalists, SciSci will be able to better identify associations discovered from models and large-scale data that have causal force to enrich their policy relevance. But experimenting on science may be the biggest challenge SciSci has yet to face. Running randomized, controlled trials that can alter outcomes for individuals or institutions of science, which are mostly supported by tax dollars, is bound to elicit criticisms and pushback (87). Hence, we expect quasi-experimental approaches to prevail in SciSci investigations in the near future. Scisci 的贡献是对科学家、机构和思想之间的关系结构的详细理解, 这是一个重要的起点, 有助于确定基本的生成过程。 这些数据驱动的努力与相关研究领域的贡献相辅相成, 如经济学(30)和科学社会学(60,86)。 因果估计是一个典型的例子, 在这个例子中, 计量经济学匹配技术要求和利用综合数据来源来模拟反事实情景(31,42)。 评估因果关系是 SciSci 未来最需要的发展之一: 许多描述性研究揭示了结构与结果之间的强烈联系, 但具体结构"导致"结果的程度尚未得到探讨。 通过与实验者建立更紧密的合作关系, SciSci 将能够更好地识别模型和大规模数据中发现的具有丰富其政策相关性的因果力量的协会。 但是科学实验可能是 SciSci 尚未面对的最大挑战。 运行随机对照试验, 可以改变个人或科学机构的结果, 这些结果大多由税收美元支持, 这必然会招致批评和反击。 因此, 我们期望在近期内准实验方法将在 SciSci 的调查中占上风。

Most SciSci research focuses on publications as primary data sources, implying that insights and findings are limited to ideas successful enough to merit publication in the first place. Yet most scientific attempts fail, sometimes spectacularly. Given that scientists fail more often than they succeed, knowing when, why, and how an idea fails is essential in our attempts to understand and improve science. Such studies could provide meaningful guidance regarding the reproducibility crisis and help us account for the file drawer problem. They could also substantially further our understanding of human imagination by revealing the total pipeline of creative activity. 大多数 SciSci 研究的重点是作为主要数据来源的出版物, 这意味着见解和研究结果仅限于那些足够成功的想法, 从一开始就值得出版。 然而, 大多数科学尝试都以失败告终, 有时甚至引人注目。 考虑到科学家失败的次数多于成功的次数, 知道什么时候, 为什么, 以及一个想法是如何失败的, 这对于我们试图理解和改进科学至关重要。 这些研究可以为重复性危机提供有意义的指导, 帮助我们解决文件抽屉问题。 他们也可以通过揭示创造性活动的总管道, 大大增进我们对人类想象力的理解。

Science often behaves like an economic system with a one-dimensional “currency” of citation counts. This creates a hierarchical system, in which the “rich-get-richer” dynamics suppress the spread of new ideas, particularly those from junior scientists and those who do not fit within the paradigms supported by specific fields. Science can be improved by broadening the number and range of performance indicators. The development of alternative metrics covering web (88) and social media (89) activity and societal impact (90) is critical in this regard. Other measurable dimensions include the information (e.g., data) that scientists share with competitors (91), the help that they offer to their peers (92), and their reliability as reviewers of their peers’ works (93). But with a profusion of metrics, more work is needed to understand what each of them does and does not capture to ensure meaningful interpretation and avoid misuse. SciSci can make an essential contribution by providing models that offer a deeper understanding of the mechanisms that govern performance indicators in science. For instance, models of the empirical patterns observed when alternative indicators (e.g., distributions of paper downloads) are used will enable us to explore their relationship with citation-based metrics (94) and to recognize manipulations. 科学常常表现得像一个一维"货币"引文计数的经济体系。 这就形成了一种等级制度, 在这种制度中,"富人致富"的动态压制了新思想的传播, 特别是那些来自初级科学家和那些不符合特定领域支持的范式的人的思想。 通过扩大业绩指标的数量和范围, 科学可以得到改进。 在这方面, 制定涵盖网络(88)和社交媒体(89)活动和社会影响的替代指标(90)在这方面至关重要。 其他可测量的方面包括科学家与竞争对手共享的信息(例如数据) , 他们给同龄人提供的帮助(92) , 以及他们作为同行作品评论者的可靠性(93)。 但是, 有了大量的指标, 我们需要做更多的工作来理解他们每个人都做了什么, 没有抓住什么, 以确保有意义的解释和避免滥用。 Scisci 可以作出重要贡献, 提供模型, 使人们更深入地了解管理科学绩效指标的机制, 从而作出重要贡献。 例如, 在使用替代指标(例如文件下载的分布)时观察到的经验模式模型将使我们能够探索它们与引文度量(94)的关系, 并识别操作。

The integration of citation-based metrics with alternative indicators will promote pluralism and enable new dimensions of productive specialization, in which scientists can be successful in different ways. Science is an ecosystem that requires not only publications, but also communicators, teachers, and detail-oriented experts. We need individuals who can ask novel, field-altering questions, as well as those who can answer them. It would benefit science if curiosity, creativity, and intellectual exchange—particularly regarding the societal implications and applications of science and technology—are better appreciated and incentivized in the future. A more pluralistic approach could reduce duplication and make science flourish for society (95). 将基于引文的衡量标准与替代指标结合起来, 将促进多元化, 并促进生产性专业化的新层面, 使科学家能够以不同的方式取得成功。 科学是一个生态系统, 不仅需要出版物, 还需要沟通者、教师和注重细节的专家。 我们需要那些能够提出新颖的、改变领域的问题的人, 以及那些能够回答问题的人。 如果好奇心、创造力和智力交流——特别是在科学和技术的社会意义和应用方面ーー未来得到更好的赞赏和激励, 科学将受益匪浅。 一个更加多元化的方法可以减少重复, 使科学为社会繁荣(95)。

An issue that SciSci seeks to address is the allocation of science funding. The current peer review system is subject to biases and inconsistencies (96). Several alternatives have been proposed, such as the random distribution of funding (97), person-directed funding that does not involve proposal preparation and review (31), opening the proposal review process to the entire online population (98), removing human reviewers altogether by allocating funds through a performance measure (99), and scientist crowd-funding (100). Scisci 试图解决的一个问题是科学资金的分配问题。 目前的同行审查制度受到偏见和不一致的影响(96)。 提出了若干备选办法, 例如随机分配资金(97)、不涉及提案编制和审查的人员指导供资(31)、向整个在线人口开放提案审查程序(98)、通过业绩计量(99)和科学家人群筹资(100)分配资金。

A critical area of future research for SciSci concerns the integration of machine learning and artificial intelligence in a way that involves machines and minds working together. These new tools portend far-reaching implications for science because machines might broaden a scientist’s perspective more than human collaborators. For instance, the self-driving vehicle is the result of a successful combination of known driving habits and information that was outside of human awareness, provided by sophisticated machine-learning techniques. Mind-machine partnerships have improved evidence-based decision-making in a wide range of health, economic, social, legal, and business problems (101–103). How can science be improved with mind-machine partnerships, and what arrangements are most productive? These questions promise to help us understand the science of the future. Scisci 未来研究的一个关键领域涉及将机器学习和人工智能结合起来, 使机器和思维协同工作。 这些新工具预示着科学的深远影响, 因为机器可能比人类合作者更能拓宽科学家的视野。 例如, 自动驾驶车辆是已知的驾驶习惯和信息成功结合的结果, 这些习惯和信息超出人类意识, 由复杂的机器学习技术提供。 心智-机器伙伴关系改善了广泛的健康、经济、社会、法律和商业问题的循证决策(101-103)。 如何通过思维机器伙伴关系来改进科学, 以及什么样的安排最有成效? 这些问题有望帮助我们理解未来的科学。

References and Notes


  • E. Garfield, Citation indexes for science; a new dimension in documentation through association of ideas. Science 122, 108–111 (1955). 10.1126/science.122.3159.108pmid:14385826FREE Full TextGoogle Scholar
  • D. J. S. Price, Little Science, Big Science (Columbia Univ. Press, 1963).Google Scholar
  • J. G. Foster, A. Rzhetsky, J. A. Evans, Tradition and innovation in scientists’ research strategies. Am. Sociol. Rev. 80, 875–908 (2015). 10.1177/0003122415601618CrossRefGoogle Scholar
  • S. Milojević, Quantifying the cognitive extent of science. J. Informetr. 9, 962–973 (2015). 10.1016/j.joi.2015.10.005Google Scholar
  • T. Kuhn, M. Perc, D. Helbing, Inheritance patterns in citation networks reveal scientific memes. Phys. Rev. X 4, 041036 (2014). 10.1103/PhysRevX.4.041036Google Scholar
  • R. Klavans, K. W. Boyack, Which type of citation analysis generates the most accurate taxonomy of scientific and technical knowledge? J. Assoc. Inf. Sci. Technol. 68, 984–998 (2016). 10.1002/asi.23734Google Scholar
  • U. Shwed, P. S. Bearman, The temporal structure of scientific consensus formation. Am. Sociol. Rev. 75, 817–840 (2010). 10.1177/0003122410388488pmid:21886269CrossRefPubMedWeb of ScienceGoogle Scholar
  • J. Bruggeman, V. A. Traag, J. Uitermark, Detecting communities through network data. Am. Sociol. Rev. 77, 1050–1063 (2012). 10.1177/0003122412463574CrossRefGoogle Scholar
  • F. Shi, J. G. Foster, J. A. Evans, Weaving the fabric of science: Dynamic network models of science’s unfolding structure. Soc. Networks 43, 73–85 (2015). 10.1016/j.socnet.2015.02.006CrossRefGoogle Scholar
  • L. M. A. Bettencourt, D. I. Kaiser, J. Kaur, Scientific discovery and topological transitions in collaboration networks. J. Informetr. 3, 210–221 (2009). 10.1016/j.joi.2009.03.001Google Scholar
  • X. Sun, J. Kaur, S. Milojević, A. Flammini, F. Menczer, Social dynamics of science. Sci. Rep. 3, 1069 (2013). 10.1038/srep01069pmid:23323212PubMedGoogle Scholar
  • T. S. Kuhn, The Essential Tension: Selected Studies in Scientific Tradition and Change (Univ. of Chicago Press, 1977).Google Scholar
  • P. Bourdieu, The specificity of the scientific field and the social conditions of the progress of reasons. Soc. Sci. Inf. (Paris) 14, 19–47 (1975). 10.1177/053901847501400602CrossRefWeb of ScienceGoogle Scholar
  • T. Jia, D. Wang, B. K. Szymanski, Quantifying patterns of research-interest evolution. Nat. Hum. Behav. 1, 0078 (2017). 10.1038/s41562-017-0078Google Scholar
  • A. Rzhetsky, J. G. Foster, I. T. Foster, J. A. Evans, Choosing experiments to accelerate collective discovery. Proc. Natl. Acad. Sci. U.S.A. 112, 14569–14574 (2015). 10.1073/pnas.1509757112pmid:26554009Abstract/FREE Full TextGoogle Scholar
  • R. Rosenthal, The file drawer problem and tolerance for null results. Psychol. Bull. 86, 638–641 (1979). 10.1037/0033-2909.86.3.638CrossRefWeb of ScienceGoogle Scholar
  • S. B. Nissen, T. Magidson, K. Gross, C. T. Bergstrom, Publication bias and the canonization of false facts. eLife 5, e21451 (2016). 10.7554/eLife.21451pmid:27995896PubMedGoogle Scholar
  • L. Yao, Y. Li, S. Ghosh, J. A. Evans, A. Rzhetsky, Health ROI as a measure of misalignment of biomedical needs and resources. Nat. Biotechnol. 33, 807–811 (2015). 10.1038/nbt.3276pmid:26252133PubMedGoogle Scholar
  • C. S. Wagner, J. D. Roessner, K. Bobb, J. T. Klein, K. W. Boyack, J. Keyton, I. Rafols, K. Börner, Approaches to understanding and measuring interdisciplinary scientific research (IDR): A review of the literature. J. Informetr. 5, 14–26 (2011). 10.1016/j.joi.2010.06.004Google Scholar
  • V. Larivière, S. Haustein, K. Börner, Long-distance interdisciplinarity leads to higher scientific impact. PLOS ONE 10, e0122565 (2015). 10.1371/journal.pone.0122565pmid:25822658CrossRefPubMedGoogle Scholar
  • K. J. Boudreau, E. C. Guinan, K. R. Lakhani, C. Riedl, Looking across and looking beyond the knowledge frontier: Intellectual distance, novelty, and resource allocation in science. Manage. Sci. 62, 2765–2783 (2016). 10.1287/mnsc.2015.2285pmid:27746512CrossRefPubMedGoogle Scholar
  • E. Leahey, J. Moody, Sociological innovation through subfield integration. Soc. Currents 1, 228–256 (2014). 10.1177/2329496514540131Google Scholar
  • A. Yegros-Yegros, I. Rafols, P. D’Este, Does interdisciplinary research lead to higher citation impact? The different effect of proximal and distal interdisciplinarity. PLOS ONE 10, e0135095 (2015). 10.1371/journal.pone.0135095pmid:26266805CrossRefPubMedGoogle Scholar
  • L. Bromham, R. Dinnage, X. Hua, Interdisciplinary research has consistently lower funding success. Nature 534, 684–687 (2016). 10.1038/nature18315pmid:27357795CrossRefPubMedGoogle Scholar
  • D. Kim, D. B. Cerigo, H. Jeong, H. Youn, Technological novelty profile and inventions future impact. EPJ Data Sci. 5, 8 (2016). 10.1140/epjds/s13688-016-0069-1Google Scholar
  • B. Uzzi, S. Mukherjee, M. Stringer, B. Jones, Atypical combinations and scientific impact. Science 342, 468–472 (2013). 10.1126/science.1240474pmid:24159044Abstract/FREE Full TextGoogle Scholar
  • J. Wang, R. Veugelers, P. Stephan, “Bias against novelty in science: A cautionary tale for users of bibliometric indicators” (NBER Working Paper No. 22180, National Bureau of Economic Research, 2016).Google Scholar
  • J. P. Walsh, Y.-N. Lee, The bureaucratization of science. Res. Policy 44, 1584–1600 (2015). 10.1016/j.respol.2015.04.010Google Scholar
  • A. M. Petersen, M. Riccaboni, H. E. Stanley, F. Pammolli, Persistence and uncertainty in the academic career. Proc. Natl. Acad. Sci. U.S.A. 109, 5213–5218 (2012). 10.1073/pnas.1121429109pmid:22431620Abstract/FREE Full TextGoogle Scholar
  • P. E. Stephan, How Economics Shapes Science (Harvard Univ. Press, 2012).Google Scholar
  • P. Azoulay, J. S. Graff Zivin, G. Manso, Incentives and creativity: Evidence from the academic life sciences. Rand J. Econ. 42, 527–554 (2011). 10.1111/j.1756-2171.2011.00140.xCrossRefWeb of ScienceGoogle Scholar
  • R. Freeman, E. Weinstein, E. Marincola, J. Rosenbaum, F. Solomon, Competition and careers in biosciences. Science 294, 2293–2294 (2001). 10.1126/science.1067477pmid:11743184FREE Full TextGoogle Scholar
  • J. A. Evans, J. G. Foster, Metaknowledge. Science 331, 721–725 (2011). 10.1126/science.1201765pmid:21311014Abstract/FREE Full TextGoogle Scholar
  • V. Larivière, C. Ni, Y. Gingras, B. Cronin, C. R. Sugimoto, Bibliometrics: Global gender disparities in science. Nature 504, 211–213 (2013). 10.1038/504211apmid:24350369CrossRefPubMedWeb of ScienceGoogle Scholar
  • S. F. Way, D. B. Larremore, A. Clauset, in Proceedings of the 25th International Conference on World Wide Web (WWW ‘16) (ACM, 2016), pp. 1169–1179.Google Scholar
  • J. Duch, X. H. T. Zeng, M. Sales-Pardo, F. Radicchi, S. Otis, T. K. Woodruff, L. A. Nunes Amaral, The possible role of resource requirements and academic career-choice risk on gender differences in publication rate and impact. PLOS ONE 7, e51332 (2012). 10.1371/journal.pone.0051332pmid:23251502CrossRefPubMedGoogle Scholar
  • J. D. West, J. Jacquet, M. M. King, S. J. Correll, C. T. Bergstrom, The role of gender in scholarly authorship. PLOS ONE 8, e66212 (2013). 10.1371/journal.pone.0066212pmid:23894278CrossRefPubMedGoogle Scholar
  • X. H. T. Zeng, J. Duch, M. Sales-Pardo, J. A. G. Moreira, F. Radicchi, H. V. Ribeiro, T. K. Woodruff, L. A. N. Amaral, Differences in collaboration patterns across discipline, career stage, and gender. PLOS Biol. 14, e1002573 (2016). 10.1371/journal.pbio.1002573pmid:27814355PubMedGoogle Scholar
  • T. J. Ley, B. H. Hamilton, The gender gap in NIH grant applications. Science 322, 1472–1474 (2008). 10.1126/science.1165878pmid:19056961PubMedGoogle Scholar
  • C. A. Moss-Racusin, J. F. Dovidio, V. L. Brescoll, M. J. Graham, J. Handelsman, Science faculty’s subtle gender biases favor male students. Proc. Natl. Acad. Sci. U.S.A. 109, 16474–16479 (2012). 10.1073/pnas.1211286109pmid:22988126Abstract/FREE Full TextGoogle Scholar
  • R. Van Noorden, Global mobility: Science on the move. Nature 490, 326–329 (2012). 10.1038/490326apmid:23075963CrossRefPubMedWeb of ScienceGoogle Scholar
  • O. A. Doria Arrieta, F. Pammolli, A. M. Petersen, Quantifying the negative impact of brain drain on the integration of European science. Sci. Adv. 3, e1602232 (2017). 10.1126/sciadv.1602232pmid:28439544FREE Full TextGoogle Scholar
  • C. Franzoni, G. Scellato, P. Stephan, The mover’s advantage: The superior performance of migrant scientists. Econ. Lett. 122, 89–93 (2014). 10.1016/j.econlet.2013.10.040CrossRefGoogle Scholar
  • C. R. Sugimoto, N. Robinson-Garcia, D. S. Murray, A. Yegros-Yegros, R. Costas, V. Larivière, Scientists have most impact when they’re free to move. Nature 550, 29–31 (2017). 10.1038/550029apmid:28980663PubMedGoogle Scholar
  • A. Clauset, S. Arbesman, D. B. Larremore, Systematic inequality and hierarchy in faculty hiring networks. Sci. Adv. 1, e1400005 (2015). 10.1126/sciadv.1400005pmid:26601125FREE Full TextGoogle Scholar
  • P. Deville, D. Wang, R. Sinatra, C. Song, V. D. Blondel, A. L. Barabási, Career on the move: Geography, stratification, and scientific impact. Sci. Rep. 4, 4770 (2014). 24759743CrossRefPubMedGoogle Scholar
  • A. M. Petersen, S. Fortunato, R. K. Pan, K. Kaski, O. Penner, A. Rungi, M. Riccaboni, H. E. Stanley, F. Pammolli, Reputation and impact in academic careers. Proc. Natl. Acad. Sci. U.S.A. 111, 15316–15321 (2014). 10.1073/pnas.1323111111pmid:25288774Abstract/FREE Full TextGoogle Scholar
  • D. K. Simonton, Creative productivity: A predictive and explanatory model of career trajectories and landmarks. Psychol. Rev. 104, 66–89 (1997). 10.1037/0033-295X.104.1.66CrossRefWeb of ScienceGoogle Scholar
  • R. Sinatra, D. Wang, P. Deville, C. Song, A.-L. Barabási, Quantifying the evolution of individual scientific impact. Science 354, aaf5239 (2016). 10.1126/science.aaf5239pmid:27811240Abstract/FREE Full TextGoogle Scholar
  • S. Wuchty, B. F. Jones, B. Uzzi, The increasing dominance of teams in production of knowledge. Science 316, 1036–1039 (2007). 10.1126/science.1136099pmid:17431139Abstract/FREE Full TextGoogle Scholar
  • N. J. Cooke, M. L. Hilton, Eds., Enhancing the Effectiveness of Team Science (National Academies Press, 2015).Google Scholar
  • V. Larivière, Y. Gingras, C. R. Sugimoto, A. Tsou, Team size matters: Collaboration and scientific impact since 1900. J. Assoc. Inf. Sci. Technol. 66, 1323–1332 (2015). 10.1002/asi.23266Google Scholar
  • L. Wu, D. Wang, J. A. Evans, Large teams have developed science and technology; small teams have disrupted it. arXiv:1709.02445 [physics.soc-ph] (7 September 2017).Google Scholar
  • B. F. Jones, The burden of knowledge and the “death of the renaissance man”: Is innovation getting harder? Rev. Econ. Stud. 76, 283–317 (2009). 10.1111/j.1467-937X.2008.00531.xCrossRefWeb of ScienceGoogle Scholar
  • S. Milojević, Principles of scientific research team formation and evolution. Proc. Natl. Acad. Sci. U.S.A. 111, 3984–3989 (2014). 10.1073/pnas.1309723111pmid:24591626Abstract/FREE Full TextGoogle Scholar
  • G. Palla, A.-L. Barabási, T. Vicsek, Quantifying social group evolution. Nature 446, 664–667 (2007). 10.1038/nature05670pmid:17410175CrossRefPubMedWeb of ScienceGoogle Scholar
  • G. J. Borjas, K. B. Doran, Which peers matter? The relative impacts of collaborators, colleagues, and competitors. Rev. Econ. Stat. 97, 1104–1117 (2015). 10.1162/REST_a_00472Google Scholar
  • P. Azoulay, J. G. Zivin, J. Wang, Superstar extinction. Q. J. Econ. 125, 549–589 (2010). 10.1162/qjec.2010.125.2.549CrossRefWeb of ScienceGoogle Scholar
  • A. M. Petersen, Quantifying the impact of weak, strong, and super ties in scientific careers. Proc. Natl. Acad. Sci. U.S.A. 112, E4671–E4680 (2015). 10.1073/pnas.1501444112pmid:26261301Abstract/FREE Full TextGoogle Scholar
  • R. K. Merton, The Matthew effect in science. Science 159, 56–63 (1968). 10.1126/science.159.3810.56Abstract/FREE Full TextGoogle Scholar
  • L. Allen, J. Scott, A. Brand, M. Hlava, M. Altman, Publishing: Credit where credit is due. Nature 508, 312–313 (2014). 10.1038/508312apmid:24745070CrossRefPubMedWeb of ScienceGoogle Scholar
  • H.-W. Shen, A.-L. Barabási, Collective credit allocation in science. Proc. Natl. Acad. Sci. U.S.A. 111, 12325–12330 (2014). 10.1073/pnas.1401992111pmid:25114238Abstract/FREE Full TextGoogle Scholar
  • L. Waltman, A review of the literature on citation impact indicators. J. Informetr. 10, 365–391 (2016). 10.1016/j.joi.2016.02.007Google Scholar
  • J. E. Hirsch, An index to quantify an individual’s scientific research output. Proc. Natl. Acad. Sci. U.S.A. 102, 16569–16572 (2005). 10.1073/pnas.0507655102pmid:16275915Abstract/FREE Full TextGoogle Scholar
  • H. F. Moed, Citation Analysis in Research Evaluation (Springer, 2010).Google Scholar
  • E. Garfield, Citation analysis as a tool in journal evaluation. Science 178, 471–479 (1972). 10.1126/science.178.4060.471pmid:5079701FREE Full TextGoogle Scholar
  • D. J. de Solla Price, Networks of scientific papers. Science 149, 510–515 (1965). 10.1126/science.149.3683.510pmid:14325149FREE Full TextGoogle Scholar
  • Q. Zhang, N. Perra, B. Gonçalves, F. Ciulla, A. Vespignani, Characterizing scientific production and consumption in physics. Sci. Rep. 3, 1640 (2013). 10.1038/srep01640pmid:23571320PubMedGoogle Scholar
  • F. Radicchi, S. Fortunato, C. Castellano, Universality of citation distributions: Toward an objective measure of scientific impact. Proc. Natl. Acad. Sci. U.S.A. 105, 17268–17272 (2008). 10.1073/pnas.0806977105pmid:18978030Abstract/FREE Full TextGoogle Scholar
  • L. Waltman, N. J. van Eck, A. F. J. van Raan, Universality of citation distributions revisited. J. Assoc. Inf. Sci. Technol. 63, 72–77 (2012). 10.1002/asi.21671Google Scholar
  • M. Golosovsky, S. Solomon, Runaway events dominate the heavy tail of citation distributions. Eur. Phys. J. Spec. Top. 205, 303–311 (2012). 10.1140/epjst/e2012-01576-4Google Scholar
  • C. Stegehuis, N. Litvak, L. Waltman, Predicting the long-term citation impact of recent publications. J. Informetr. 9, 642–657 (2015). 10.1016/j.joi.2015.06.005CrossRefGoogle Scholar
  • M. Thelwall, The discretised lognormal and hooked power law distributions for complete citation data: Best options for modelling and regression. J. Informetr. 10, 336–346 (2016). 10.1016/j.joi.2015.12.007Google Scholar
  • D. de Solla Price, A general theory of bibliometric and other cumulative advantage processes. J. Am. Soc. Inf. Sci. 27, 292–306 (1976). 10.1002/asi.4630270505CrossRefWeb of ScienceGoogle Scholar
  • A.-L. Barabási, R. Albert, Emergence of scaling in random networks. Science 286, 509–512 (1999). 10.1126/science.286.5439.509pmid:10521342Abstract/FREE Full TextGoogle Scholar
  • P. D. B. Parolo, R. K. Pan, R. Ghosh, B. A. Huberman, K. Kaski, S. Fortunato, Attention decay in science. J. Informetr. 9, 734–745 (2015). 10.1016/j.joi.2015.07.006Google Scholar
  • D. Wang, C. Song, A.-L. Barabási, Quantifying long-term scientific impact. Science 342, 127–132 (2013). 10.1126/science.1237825pmid:24092745Abstract/FREE Full TextGoogle Scholar
  • Y.-H. Eom, S. Fortunato, Characterizing and modeling citation dynamics. PLOS ONE 6, e24926 (2011). 10.1371/journal.pone.0024926pmid:21966387CrossRefPubMedGoogle Scholar
  • M. Golosovsky, S. Solomon, Stochastic dynamical model of a growing citation network based on a self-exciting point process. Phys. Rev. Lett. 109, 098701 (2012). 10.1103/PhysRevLett.109.098701pmid:23002894CrossRefPubMedGoogle Scholar
  • A. F. J. van Raan, Sleeping Beauties in science. Scientometrics 59, 467–472 (2004). 10.1023/B:SCIE.0000018543.82441.f1CrossRefWeb of ScienceGoogle Scholar
  • Q. Ke, E. Ferrara, F. Radicchi, A. Flammini, Defining and identifying Sleeping Beauties in science. Proc. Natl. Acad. Sci. U.S.A. 112, 7426–7431 (2015). 10.1073/pnas.1424329112pmid:26015563Abstract/FREE Full TextGoogle Scholar
  • I. Tahamtan, A. Safipour Afshar, K. Ahamdzadeh, Factors affecting number of citations: A comprehensive review of the literature. Scientometrics 107, 1195–1225 (2016). 10.1007/s11192-016-1889-2Google Scholar
  • J. E. Hirsch, Does the h index have predictive power? Proc. Natl. Acad. Sci. U.S.A. 104, 19193–19198 (2007). 10.1073/pnas.0707962104pmid:18040045Abstract/FREE Full TextGoogle Scholar
  • D. E. Acuna, S. Allesina, K. P. Kording, Future impact: Predicting scientific success. Nature 489, 201–202 (2012). 10.1038/489201apmid:22972278CrossRefPubMedWeb of ScienceGoogle Scholar
  • O. Penner, R. K. Pan, A. M. Petersen, K. Kaski, S. Fortunato, On the predictability of future impact in science. Sci. Rep. 3, 3052 (2013). 10.1038/srep03052pmid:24165898CrossRefPubMedGoogle Scholar
  • J. R. Cole, H. Zuckerman, in The Idea of Social Structure: Papers in Honor of Robert K. Merton, L. A. Coser, Ed. (Harcourt Brace Jovanovich, 1975), pp. 139–174.Google Scholar
  • P. Azoulay, Research efficiency: Turn the scientific method on ourselves. Nature 484, 31–32 (2012). 10.1038/484031apmid:22481340CrossRefPubMedGoogle Scholar
  • M. Thelwall, K. Kousha, Web indicators for research evaluation. Part 1: Citations and links to academic articles from the Web. Prof. Inf. 24, 587–606 (2015). 10.3145/epi.2015.sep.08Google Scholar
  • M. Thelwall, K. Kousha, Web indicators for research evaluation. Part 2: Social media metrics. Prof. Inf. 24, 607–620 (2015). 10.3145/epi.2015.sep.09Google Scholar
  • L. Bornmann, What is societal impact of research and how can it be assessed? A literature survey. Adv. Inf. Sci. 64, 217–233 (2013).Google Scholar
  • C. Haeussler, L. Jiang, J. Thursby, M. Thursby, Specific and general information sharing among competing academic researchers. Res. Policy 43, 465–475 (2014). 10.1016/j.respol.2013.08.017Google Scholar
  • A. Oettl, Sociology: Honour the helpful. Nature 489, 496–497 (2012). 10.1038/489496apmid:23018949CrossRefPubMedGoogle Scholar
  • S. Ravindran, “Getting credit for peer review,” Science, 8 February 2016; Scholar
  • R. Costas, Z. Zahedi, P. Wouters, Do “altmetrics” correlate with citations? Extensive comparison of altmetric indicators with citations from a multidisciplinary perspective. J. Assoc. Inf. Sci. Technol. 66, 2003–2019 (2015). 10.1002/asi.23309Google Scholar
  • A. Clauset, D. B. Larremore, R. Sinatra, Data-driven predictions in the science of science. Science 355, 477–480 (2017). 10.1126/science.aal4217Abstract/FREE Full TextGoogle Scholar
  • S. Wessely, Peer review of grant applications: What do we know? Lancet 352, 301–305 (1998). 10.1016/S0140-6736(97)11129-1pmid:9690424CrossRefPubMedWeb of ScienceGoogle Scholar
  • N. Geard, J. Noble, paper presented at the 3rd World Congress on Social Simulation, Kassel, Germany, 6 to 9 September 2010.Google Scholar
  • Calm in a crisis. Nature 468, 1002 (2010). 10.1038/4681002apmid:21170024PubMedGoogle Scholar
  • R. Roy, Funding science: The real defects of peer review and an alternative to it. Sci. Technol. Human Values 10, 73–81 (1985). 10.1177/016224398501000309CrossRefGoogle Scholar
  • J. Bollen, D. Crandall, D. Junk, Y. Ding, K. Börner, An efficient system to fund science: From proposal review to peer-to-peer distributions. Scientometrics 110, 521–528 (2017). 10.1007/s11192-016-2110-3Google Scholar
  • M. S. Kohn, J. Sun, S. Knoop, A. Shabo, B. Carmeli, D. Sow, T. Syed-Mahmood, W. Rapp, IBM’s health analytics and clinical decision support. Yearb. Med. Inform. 9, 154–162 (2014). 10.15265/IY-2014-0002pmid:25123736CrossRefPubMedGoogle Scholar
  • J. Kleinberg, H. Lakkaraju, J. Leskovec, J. Ludwig, S. Mullainathan, “Human decisions and machine predictions” (National Bureau of Economic Research, 2017).Google Scholar
  • B. Liu, R. Govindan, B. Uzzi, Do emotions expressed online correlate with actual changes in decision-making?: The case of stock day traders. PLOS ONE 11, e0144945 (2016). 10.1371/journal.pone.0144945pmid:26765539PubMedGoogle Scholar