Flat Teams Drive Scientific Innovation

2025-04-27 0 0 93.64KB 4 页 10玖币

侵权投诉

arXiv:2210.05852v1 [cs.DL] 12 Oct 2022

QUANTIFYING HIERARCHY IN SCIENTIFIC TEAMS

A PREPRINT

Fengli Xu

Knowledge Lab, Department of Sociology

University of Chicago

Chicago, IL

fenglixu@uchicago.edu

Lingfei Wu

School of Computing and Information

University of Pittsburgh

Pittsburgh, PA

liw105@pitt.edu

James A. Evans

Knowledge Lab, Department of Sociology

University of Chicago

Chicago, IL

jevans@uchicago.edu

October 13, 2022

ABSTRACT

This paper provides a detailed description of the data collection and machine learning model used

in our recent PNAS paper "Flat Teams Drive Scientiﬁc Innovation" Xu et al. [2022a]. Here, we

discuss how the features of scientiﬁc publication can be used to estimate the implicit hierarchy in

the corresponding author teams. Besides, we also describe the method of evaluating the impact of

team hierarchy on scientiﬁc outputs. More details will be updated in this article continuously. Raw

data and Readme document can be accessed in this GitHub repository Xu et al. [2022b].

Keywords Teams ·Innovation ·Science of science ·Productivity ·Novelty ·Disruption

1 Estimating the Impact of Team Hierarchy (L-ratio) on Scientiﬁc Output

To evaluate the impact of team hierarchy (L-ratio) on the innovation performance of individual scientists, we select

scientists publishing two or more papers and perform author and ﬁeld ﬁxed-effect regressions to normalize the perfor-

mance differences between authors and ﬁelds of study as deﬁned in Microsoft Academic Graph mag [2019-03-22]. We

include L-ratio as an independent variable to predict six dependent variables, including novelty, developmental index,

the productivity of lead authors, the productivity of support authors, short-term citation impact, and long-term citation

impact. To control for possible confounders, we also include team size Wu et al. [2019], the mean, standard deviation

and max value of career age Blau and Weinberg [2017], whether the work is supported by funding agencies, and the

number and award amount of received funding as independent variables in predicting each dependent variable. We

ﬁnd L-ratio continues to be statistically signiﬁcant across all regressions. L-ratio explains the most additional variance

in predicting the increased developmental index (167%), increased short-term citations (121%), and decreased novelty

(17%). The estimated regression coefﬁcients also allow us to compare the effect of different variables. This suggests,

for example, that changing team structure (engaging a support author as a lead author) may be more effective than

changing team size in maximizing novelty.

2 Quantifying the Novelty of Research Papers

The novelty metric (Fig.2a) quantiﬁes to what extent a paper links topic keywords that rarely appear together. Our

metric is designed to extend the Uzzi score of reference novelty. Brian Uzzi and colleagues created a prominent

score that captures how a paper deviates from the norm of science by building on “atypical” references, where a pair

文档加载中……请稍候！
如果长时间未打开，您也可以点击刷新试试。

下载文档到电脑，查找使用更方便

10 玖币 0人已下载

立即下载

摘要：

arXiv:2210.05852v1[cs.DL]12Oct2022QUANTIFYINGHIERARCHYINSCIENTIFICTEAMSAPREPRINTFengliXuKnowledgeLab,DepartmentofSociologyUniversityofChicagoChicago,ILfenglixu@uchicago.eduLingfeiWuSchoolofComputingandInformationUniversityofPittsburghPittsburgh,PAliw105@pitt.eduJamesA.EvansKnowledgeLab,DepartmentofS...

展开>> 收起<<

Flat Teams Drive Scientific Innovation.pdf

共4页,预览1页

还剩页未读，继续阅读

声明：本站为文档C2C交易模式，即用户上传的文档直接被用户下载，本站只是中间服务平台，本站所有文档下载所得的收益归上传人(含作者)所有。玖贝云文库仅提供信息存储空间，仅对用户上传内容的表现方式做保护处理，对上载内容本身不做任何修改或编辑。若文档所含内容侵犯了您的版权或隐私，请立即通知玖贝云文库，我们立即给予删除！

Flat Teams Drive Scientific Innovation

相关推荐

开通VIP享超值会员特权

作者详情

相关内容

热门标签

举报选择: