Performance evaluation of “Open Bidding for Selecting the Best Candidates” research teams based on big data analysis

As the global technological revolution and industrial transformation accelerate, scientific innovation and societal challenges have become increasingly complex, leading to unprecedented interdisciplinary integration and collaboration levels. Individual researchers in single fields face growing difficulties in solving major technological problems and achieving breakthrough innovations. As a result, interdisciplinary, cross-organizational research teams based on collaborative efforts have gradually become mainstream (Chen et al., 2022; Huang et al., 2022). A research team centered around a team leader and focused on scientific and technological research and development is a group of researchers with complementary knowledge and skills who share a common vision and goals, take mutual responsibility, and maintain relatively close and stable cooperation on concentrated research areas and scientific problems of mutual concern (Chen and Yang, 2002; Pang et al., 2011; Yu et al., 2018). As the fundamental organizational unit of research institutions and the primary driving force behind academic innovation, research teams play an indispensable role in solving complex scientific problems, accelerating the commercialization of scientific results, and promoting economic and social development.

Current research on scientific teams primarily relies on systematically analyzing research data. Co-authors of the same academic paper are typically identified as a research team (Zeng et al., 2017), a definition that provides a foundational framework for quantitative analysis. With the widespread availability of big data in scientific research, an increasing volume of data on funding, academic output, researcher collaboration networks, and citation patterns has emerged, offering unprecedented opportunities to analyze the structure and dynamics of the scientific system. Data-driven analytical methods allow for an in-depth exploration of scientific big data, offering ways to quantify the interactions between fundamental scientific units across different temporal and spatial scales, thereby providing theoretical support for advancing scientific progress (Fortunato et al., 2018).

As a result, the identification and performance evaluation of research teams based on research data has become a focal point for academia and policymakers.

At present, scholars at home and abroad have achieved remarkable results in research team identification, and these studies are mainly based on the disciplinary, institutional and journal perspectives to identify research teams.

Due to differences in the characteristics and research paradigms of various disciplines, the development of each field varies to some extent. As a result, research teams’ structure, size, and performance evaluation differ across disciplines. Consequently, research findings on scientific teams often exhibit diverse traits depending on the disciplinary perspective, and it is not appropriate to simply apply the results from one discipline to another. Instead, team identification metrics should be tailored to each discipline’s specific context, and research teams’ evolutionary patterns should be explored accordingly (Li et al., 2023).

Chen et al. (2015), based on a search of the “management” discipline in the Chinese Social Sciences Citation Index (CSSCI) database for literature from 2004 to 2013, employed social network analysis to construct an author co-authorship network. This led to the identification of representative academic groups in the field of innovation management in China. Research revealed that most of these groups originated within the same university, while the largest sub-connected network formed a cross-regional, cross-institutional, and interdisciplinary collaborative network. Similarly, Li and Zhang (2019) reviewed professional journals and relevant conference papers. They conducted expert interviews to summarize various research teams’ research directions, methods, and future trends in the ABC and DEF fields. Zhang et al. (2020), Bi and Su (2017), and Dang et al. (2023) utilized the SSCI international academic journal database and 12 key domestic journals, applying bibliometric methods and social network analysis to study core scholars and academic collaborations in the field of tourism research in China. Nian et al. (2023) analyzed publication volume and author collaboration networks in evidence-based social science research using the Web of Science, the China Science Citation Database, and the CSSCI database. Yu et al. (2020), based on data from the Web of Science on AI-related papers from 2009 to 2018, used fractional counting to construct an author collaboration network, selecting six indicators such as publication volume, citation count, and betweenness centrality to identify leading teams in the AI field. They further conducted a comparative analysis of these teams through paper data and empirical surveys. Lastly, Lv and Tan (2022) applied an improved Louvain algorithm to analyze NIH-funded research papers, conference papers, and reviews published in the Scopus database from 2012 to 2021 in critical care medicine. By assessing the degree of close collaboration within the co-authorship networks of researchers in this interdisciplinary field, they identified virtual research teams and conducted case studies of specific research groups.

Some scholars have focused on explicit collaboration networks to identify innovative research teams from the perspective of specific institutions using research data from academic publications. For instance, Wei et al. (2013), taking the Institute of Scientific and Technical Information of China as an example, utilized the full-text journal database of CNKI to explore two methods for selecting team members: one based on explicit collaboration networks and the other on implicit collaboration networks. Their study demonstrated that explicit collaboration networks are more suitable for selecting team members centered around individuals, while implicit collaboration networks are better suited for team selection focused on research topics. In the actual team formation process, both methods can be combined. Zhang et al. (2016), using all SCI papers co-authored by the Chinese Academy of Sciences before 2015 as a sample, applied social network analysis to explore the characteristics of its industry-university-research collaboration network. Through regression analysis, they examined how these features impact basic research performance. The study found that the collaboration intensity between the Chinese Academy of Sciences and enterprises was significantly lower than that with universities, with stronger collaboration with domestic universities compared to foreign ones, while the intensity of collaboration with foreign enterprises was higher than that with domestic enterprises.

Some scholars focus on papers published in specific journals, using publication data to identify research teams in particular fields and uncover the development of scientific collaboration within those fields. For example, Zhang and Yan (2010) conducted a social network analysis of authors who published more than two papers in the journal Management Review between 2004 and 2008, identifying several research groups in management science. They conducted an in-depth study of these collaboration networks’ structure, characteristics, and competitiveness. Their research revealed that collaboration networks in China’s management science field are generally sparse, with a few scholars playing pivotal roles in the overall network structure, and the competitiveness of these scholars is closely related to their academic standing within subnetworks. Similarly, Yuan and Xun (2018), using papers published in four comparative education journals—Comparative Education Review, Global Education Outlook, Foreign Education Studies, and Foreign Primary and Secondary Education—from 2000 to 2016 as a sample, employed co-citation analysis and multivariate statistical methods to cluster and classify the primary academic teams in the field of comparative education. They identified 35 core scholars and divided them into six main academic groups. Their study found that the density and clustering of the co-citation network of core scholars exhibit characteristics of a small-world effect, reflecting smooth communication and active interaction among academic communities in China’s comparative education research field, where academic discussions have become a norm. Jiang et al. (2020), using the collaboration among authors of papers published in the Journal of Sport Science from 1998 to 2018 as a sample, identified five representative research teams through social network analysis based on core authors who published more than 10 papers. By analyzing the network density, centrality, core-periphery structure, and cohesive subgroups of the co-authorship network, the study found that collaboration within sports-related research teams is influenced by factors such as geographic location, research direction, mentor-mentee relationships, institutional affiliation, and educational background. The research revealed minimal factionalism, with small groups focused on internal information exchange but lacking cross-group communication.

By combining the relevant literature, this paper finds that academic research on the evaluation of research team performance is mainly carried out in the following seven dimensions.

Research investment intensity refers to the total amount and concentration of resources—such as funding, manpower, equipment, time, and knowledge—allocated by a research team to scientific projects within a specific timeframe (Wang, 2013). This metric is significant for assessing research teams’ development potential, competitiveness, and management efficiency. By enhancing investment intensity, teams can improve their research capabilities and competitiveness, leading to more substantial research outcomes. Existing studies often select secondary indicators under the research investment intensity metric, including material, time, and personnel investment. Specifically, Zhang (2014), Zhu and Yang (2013) identified funding from industry-university collaborations, government research grants, and social sponsorships as evaluation metrics for team investment. They view these resources as the financial backbone of the research team, essential for carrying out various tasks and indicating whether the team has secured strong support. Wang (2013) categorized secondary indicators under the “team investment” dimension into material investment, time investment, and human resource investment. In particular, material investment is measured by funding sources, expenditure, equipment consumption, and library resources; time investment is represented by the duration required to complete projects; and the number and structure of personnel characterizes human resource investment. Data on research investment intensity is typically sourced from detailed surveys of research teams, complemented by comprehensive analyses of other relevant data sources.

Research output capability refers to the level researchers demonstrate in their scientific achievements, primarily characterized by the quantity and quality of their research outputs (Zhao and Ma, 2018). Currently, many studies use publications as the core data for measuring research output levels, thereby constructing a multidimensional evaluation system that includes indicators of output quantity, quality, impact, innovation, and development potential (Zhao et al., 2014). The specific evaluation indicators encompass metrics reflecting quantity, such as the number of papers published and the annual growth rate, as well as quality assessment metrics (e.g. JCR rankings), average citation counts, impact factors, the volume of publications in core journals, the proportion of papers in top-tier journals, h-index, and the ratio of highly cited papers (Wang and Yang, 2012; Wei, 2014; Ye et al., 2015; Yu and Zhang, 2019; Zhong, 2020).

Specifically, Lou (1995) utilized bibliometric data to comprehensively assess research institutions’ effectiveness and output capability through multiple dimensions, including publication output rate, award points, international indexing status, and citation frequency. Cui (2011) further proposed that the funding status of papers should be included as an evaluation metric when analyzing the research level of libraries. Moreover, some scholars have expanded beyond traditional paper output assessments to incorporate diverse research outputs into the evaluation system of academic research levels in universities or research institutions, including patents, monographs, research projects, and funding. These newly added evaluation dimensions encompass but are not limited to the practical application and transformation of research results, the development of software and standards, awards at various levels (national and provincial), the number of granted patents, and the status of research project approvals, thus constructing a more comprehensive and multidimensional framework for evaluating research output (Wang and Sun, 2017).

A scientific research team’s research content and themes are central components of its activities, directly reflecting the team’s research focus and, more profoundly, indicating the breadth and depth of knowledge exchange within the team. This aspect is indispensable for evaluating the team’s innovative potential and academic influence. When a research team exhibits minimal changes in its research direction over different periods, it often signifies a sustained commitment to a specific field, resulting in a solid professional foundation and academic accumulation. To quantify and analyze this characteristic, Wang and Zhang (2014) proposed a measurement method based on the “degree of variation in research themes” to assess the continuity of research directions in university research teams. Specifically, this method constructs a knowledge network based on keywords from the papers and academic works published by team members, dynamically monitoring and comparing the structural changes of the knowledge network over different periods. Chen et al. (2023) developed a co-occurrence matrix that includes all paper keywords and their publication years. They evaluated the continuity of research themes by calculating the proportion of keywords that repeat from one year to the previous year over various periods.

Modern scientific research has entered the “Big Science” era, with grand research objectives, high-intensity investments, and interdisciplinary collaboration (Chen and Sheng, 2011). In this context, scientists from diverse fields actively form multidisciplinary research teams to tackle significant social and economic issues through collaborative research. This form of interdisciplinary cooperation not only facilitates the cross-pollination of knowledge and innovation but also serves as a key driving force behind groundbreaking scientific discoveries and technological advancements (Zhai et al., 2023). By pooling expertise and resources from various disciplines, scientific communities can take a more comprehensive view of problems and propose more forward-thinking and integrated solutions, thereby fostering a deep integration of scientific progress and societal development (Qian, 2007; Liu and Chen, 2007). Consequently, some scholars have identified interdisciplinary integration capability as essential in evaluating research teams. Regarding specific metrics, Stirling (2007) assessed the “knowledge integration level” in cross-disciplinary research through three dimensions: the richness of disciplines, the uniformity of discipline distribution, and the degree of disciplinary difference. Zhuang (2020), in constructing an evaluation system for university research teams, evaluated interdisciplinary integration capability by considering the level of collaboration in academic activities such as publishing papers, applying for patents, and writing monographs. Liao et al. (2021) assessed a research team’s interdisciplinary integration capability using a “team interdisciplinarity” metric based on the disciplines associated with their members’ institutions.

By assessing the collaboration density of research project teams, one can gain valuable insights into the cooperative dynamics among team members, which is significantly informative for optimizing personnel configuration and enhancing efficiency. Numerous studies have shown a positive correlation between the degree of research collaboration and the research performance of entities (Yang and Li, 2019; Zhang et al., 2019; Ma and Li, 2020). Using complex network theory, one method for evaluating team collaboration density is to construct a network diagram of paper collaborations among team members. Specifically, the structural characteristics of the co-authorship network—including node (member) density, edge distribution (establishment of collaborative relationships), and network clustering coefficients—serve as important indicators for assessing the level of collaboration among team members. A comprehensive analysis of these indicators can reveal the diversity and efficiency of collaborative patterns within the team, thereby distinguishing significant differences between high and low-collaboration teams. The co-authorship network in highly efficient research teams typically exhibits a more compact structure, with tighter and more diverse connections among members. Collaborative behaviors display a high degree of synergy and organic integration, greatly facilitating knowledge sharing, innovative thinking, and the generation of research outputs. Qi (2013) developed a formula for “collaboration network cohesion” based on complex network analysis methods to analyze the tightness of team collaboration and its evolution over time. Chen (2023) employed four indicators—“team member collaboration degree”, “team member co-authorship rate”, “network density”, and “clustering index”—to measure the cohesion of university research teams while constructing an evaluation index system for high-level research teams in the humanities and social sciences in Jiangxi Province.

Research teams should not only focus on their research outputs but also prioritize their development and growth. The growth capability of a research team reflects its ability to adapt and evolve within a constantly changing research environment, directly impacting the team’s long-term development and innovative potential (Zhang, 2014). Existing studies often measure the growth capability of research teams using indicators such as “talent development”, “research platform establishment”, “academic exchange”, and “social reputation.” For instance, Zhu et al. (2009) selected four secondary indicators—talent development, research platform establishment, academic exchange, and team social reputation—to assess the learning and growth capabilities of university technology innovation teams. Similarly, Shan et al. (2013) employed a mixed qualitative and quantitative approach to investigate the growth capability of agricultural technology innovation teams through aspects such as project engagement, research platform development, talent cultivation, and academic exchange. Building on indicators related to talent development and research platforms, Chen et al. (2022) further incorporated research funding and interdisciplinary collaboration to assess the sustainable development capability of university research teams.

Research team effectiveness primarily refers to the social impact and economic value of research outputs. A greater effectiveness indicates a larger influence of the research team, reflecting, to some extent, the quality of research results and demonstrating the practical application value of research outputs in social and economic contexts (Wang, 2013; Shen and Feng, 2015). Effectiveness typically encompasses social benefits, economic benefits, and talent development. Social benefits pertain to the positive impact of research results on society, including the team’s social recognition and reputation. Economic benefits focus on the commercial value generated from the transformation of research outcomes and the prospects of those results. Talent development reflects the team’s effectiveness in cultivating research talent, as evidenced by the achievements of team members, their professional title evaluations, and the mentoring of master’s and doctoral students. Zhang and He (2010) measured the performance of university innovation teams by selecting secondary indicators such as per capita output rate, asset output rate (in ten thousand yuan), researcher training, and outcome transformation. Zhang and Han (2015), in constructing an evaluation index system for the sustainable innovation capability of university technology innovation teams, utilized surveys to measure the social impact and economic value of core research outcomes as two qualitative indicators of research team effectiveness.

A review of the literature on research team identification reveals significant diversity in existing studies, primarily framed around three perspectives: disciplines, institutions, and journals. First, researchers generally recognize the uniqueness of research teams across different disciplines, emphasizing the need to customize identification indicators based on disciplinary characteristics. Second, methods such as social network analysis and bibliometrics are widely employed to identify research teams through the construction of co-authorship networks and analysis of publication data. Third, the diverse data sources—including domestic and international databases, specialized journals, and conference proceedings—ensure the breadth and depth of the research.

However, existing studies also exhibit several shortcomings: First, the treatment of inter-disciplinary differences remains insufficiently nuanced, sometimes overlooking specific needs of certain disciplines. Second, limitations in data sources and methodologies may affect the comprehensiveness and accuracy of identification results. Third, there is inadequate exploration of internal collaboration mechanisms within research teams and their interactions with external environments. Fourth, there is a scarcity of studies examining the discovery and selection mechanisms of research teams based on specific institutional arrangements (such as research funding policies and project management mechanisms).

In light of these shortcomings, particularly the lack of research focused on the identification and selection mechanisms of research teams within specific institutional contexts, this paper aims to conduct an empirical analysis of research teams under the “Open Bidding for Selecting the Best Candidates” policy framework.

In the field of research team performance evaluation, numerous studies have explored various perspectives, including input-output and internal-external viewpoints. These studies assess performance across dimensions such as research input intensity, research output capability, continuity of research direction, interdisciplinary integration, collaboration intensity, growth capacity, and overall benefits of research teams. They utilize both quantitative and qualitative indicators, including material input, human resources, publication volume, citation rates, knowledge integration, and social benefits. However, when analyzing and evaluating multiple research teams, the significant differences in team types and their environmental contexts often pose challenges for the existing evaluation frameworks during horizontal comparisons, revealing certain limitations.

From the perspective of data acquisition, the collection and quantification of qualitative indicators—such as material input, social benefits, research platform development, talent exchange, and team reputation—are particularly complex and difficult to standardize. This complexity largely restricts the broad applicability and depth of analysis of the evaluation system. Furthermore, the evaluation frameworks established in existing research are often tailored to specific goals or fields. This implies that when designing evaluation models, it is necessary to consult relevant experts based on the characteristics of the target domain and adjust weight settings accordingly, resulting in a lack of generalizability for the evaluation system.

The primary goal of research teams engaged in innovative scientific activities is to produce research outcomes. In both basic and applied foundational research, these outcomes are often presented in the form of academic papers. Analyzing research papers can provide insights into a team’s innovative capacity and performance. Therefore, this study further narrows its focus when identifying and evaluating research teams, particularly from the perspective of publication output. Building on existing research, we randomly selected 33 leaders of research teams from universities, based on the data from the “Open Bidding for Selecting the Best Candidates” policy text database, who published papers in CSSCI-indexed and Peking University core journals between 2000 and February 2024 on CNKI. We employed co-authorship network clustering analysis to identify the research teams led by these 33 leaders.

When evaluating the performance of research teams, indicators such as research input intensity and team benefits typically require data obtained through surveys, which are difficult to quantify and can be significantly influenced by human factors. Additionally, the social and economic benefits of research teams are not the direct objectives of scientific research, nor are they the focus of this evaluation; thus, these indicators are excluded from the performance evaluation framework for Beijing’s “Open Bidding for Selecting the Best Candidates” research teams.

Furthermore, since this study examines the publication patterns of research team leaders over the past 24 years, the publication volume and other metrics under the research output capacity dimension, as well as indicators related to continuity of research direction, also reflect the growth potential and future development capacity of the research teams to some extent. Given that the indicators used to measure growth capacity—such as talent development, research platform establishment, academic exchange, and social reputation—are predominantly qualitative in nature, data acquisition is challenging and standardization is difficult. Therefore, this study does not include growth capacity indicators in the evaluation framework.

Following the selection criteria of scientific validity, comparability, representativeness, systematic approach, and quantifiability, this paper constructs a performance evaluation indicator system for research teams, which comprises four primary indicators: research output capacity, continuity of research direction, interdisciplinary integration, and collaboration intensity, along with several secondary indicators such as publication volume of the leaders, as shown in Table 1.

Specifically, under the primary indicator of research output capacity, this study considers the operational feasibility and general applicability of the research. It selects two secondary indicators—“publication volume of team leaders” and “proportion of papers in top-tier journals”—to assess the research output capability of the “Open Bidding for Selecting the Best Candidates” teams from the perspectives of both quantity and quality of publications.

Publication volume is a key indicator of a researcher’s and a research team’s scholarly level and academic competence (Wang et al., 2018a, b). By examining publication volume, one can intuitively gauge the number of research outputs produced by a team in a particular field, reflecting their activity and contributions in that area. The “publication volume of team leaders” refers to the total number of papers published by leaders of the “Open Bidding for Selecting the Best Candidates” teams on CNKI from 2000 to 2024. This study further analyzes the dynamic trends in publication volume in the five years before and after the leaders receive their academic titles, in order to evaluate changes in their scholarly output.

In alignment with existing research, such as studies by Chen et al. (2016) and Feng (2014), which highlight the number of papers indexed in SCI, ISTP, and CSSCI as important indicators of a scientist’s innovative capability, this study also incorporates “proportion of papers in top-tier journals” to measure the quality of research output. This is expressed as the ratio of papers published in leading Chinese journals to the total number of papers. Papers published in top-tier journals are characterized by high levels of innovation; therefore, a higher proportion indicates greater productivity and scholarly impact from the research team.

Under the primary indicator of research direction continuity, the secondary indicator selected is “whether the themes of research outcomes have changed”. The specific measurement method involves analyzing the keywords and abstracts of papers published by the team leaders to determine if there has been a shift in research themes in the five years before and after the leaders received their academic titles. This evaluation indicator assesses the research direction of the scientific team and its continuity, indicating whether the team has maintained a consistent focus on the same research area over different time spans. A high level of continuity suggests that the team’s research focus is stable, potentially due to their ongoing efforts to deepen and expand upon previous research. Conversely, low continuity may indicate that the team’s research interests are fragmented or lack coherence, suggesting a need for more strategic planning regarding the team’s research direction.

Under the primary indicator of interdisciplinary integration ability, the secondary indicator chosen is “degree of interdisciplinary collaboration”. The specific measurement method involves analyzing the co-occurrence of keywords or themes in the team’s published papers to assess the interconnections between different disciplines. If a team’s publications include keywords or themes from multiple disciplines with a high frequency of co-occurrence, this indicates a strong capability for interdisciplinary integration.

Under the primary indicator of team collaboration intensity, special emphasis is placed on the role of the leader in the “Open Bidding for Selecting the Best Candidates” teams, as they are central to the group dynamics. Therefore, this indicator focuses on the leader’s ability to guide the research efforts of team members, measured through the collaboration on co-authored papers. Specifically, the secondary indicator selected is “co-authorship rate”, defined as the proportion of papers co-authored by the leader and team members relative to the total number of papers published. The co-authorship rate reflects the leader’s academic guidance and capacity for enhancing the skills of team members; a higher co-authorship frequency suggests closer collaboration among team members, allowing them to effectively leverage their respective expertise to produce a greater quantity and higher quality of research outcomes. Furthermore, a higher frequency of co-authorship indicates a supportive research environment that fosters knowledge sharing, communication, and learning, which in turn contributes to the academic growth of team members and enhances the overall professional reputation of the team.

Universities host a multitude of high-level talents, many of whom play a central role in “Open Bidding for Selecting the Best Candidates” research teams. This is especially true for experts and scholars from top institutions such as those categorized as “Double First-Class,” “985,” and “211,” who form the backbone of these teams (Zeng et al., 2023). Consequently, this study focuses on team leaders from universities involved in “Open Bidding for Selecting the Best Candidates” projects, aiming to delve into the composition and performance of these academic leaders’ research teams.

To this end, a selection of 33 university-based “Open Bidding for Selecting the Best Candidates” research team leaders was made from the policy text database as the analytical sample. Their publication records from 2000 to February 2024 in CSSCI (Chinese Social Sciences Citation Index) and Peking University core journals were collected using the CNKI (China National Knowledge Infrastructure) database, allowing for the identification of their research teams based on co-authorship patterns.

The data for this study primarily comes from CNKI, a globally leading and comprehensive Chinese journal database. CNKI provides real-time updates of academic outputs and encompasses a variety of resources, including academic journals, theses, conference proceedings, newspapers, yearbooks, and books. By integrating knowledge and information resources, CNKI has established a platform for the exchange of networked knowledge resources, promoting the sharing of social resources and facilitating digital learning. The academic resources available in this database cover various fields, including basic sciences, medical and health sciences, agricultural technology, philosophy and humanities, and economics and management. Given that this study focuses on examining the co-authorship patterns of “Open Bidding for Selecting the Best Candidates” research teams across different disciplines, CNKI was chosen as the primary data source.

The academic community primarily employs four methods for identifying research teams: questionnaire surveys, association rule algorithms, clustering analysis, and social network analysis.

Field-based questionnaire surveys are the most common method for identifying research teams. Researchers collect relevant data through on-site investigations, utilizing questionnaires, expert discussions, and other formats to discover and identify research teams, as well as to analyze collaboration patterns, evolutionary characteristics, member traits, and innovative performance within those teams. This method is primarily used to identify and discover tangible teams, whose members are usually geographically proximate, often belong to the same organization, and interact through face-to-face communication, leading to relatively clear organizational boundaries (Li, 2023).

The questionnaire survey method, based on field research, is the most common approach for identifying research teams. Researchers collect relevant data through surveys and expert discussions, allowing them to discover and analyze collaboration patterns, evolutionary characteristics, member traits, and innovation performance within research teams. This method is particularly useful for identifying tangible teams, as their members are usually geographically close, often belonging to the same organization, and interact face-to-face, making organizational boundaries relatively clear (Li, 2023).

For instance, Li (2012) distributed questionnaires to traditional Chinese medicine institutions and hospitals across 31 provinces to analyze the development of research teams in major disease studies. This study compared the strengths and weaknesses of research teams focused on different diseases. Similarly, Jing and Hu (2014) conducted paired surveys with 150 research teams from six universities nationwide, exploring how the relationship between team leaders and members affects team performance. Their findings indicated that the leader’s role and the professional respect they command significantly influence emotional exchanges between members and leaders.

Additionally, Gu and Wang (2015) issued questionnaires to research and development teams in industries like mobile communications and aerospace, finding that the three dimensions of social capital—structural, cognitive, and relational—significantly impacted innovation performance. Zeng and Li (2019) utilized a combination of on-site and mail surveys to investigate research teams across various universities and research institutes, focusing on the misalignment phenomena in technology transfer between academia and industry.

The questionnaire survey method relies on project initiation information or institutional introductions to identify research teams, emphasizing geographic proximity and resource sharing. While this method is straightforward, accurate, and easily understood, making it effective for data collection, it is also time-consuming and costly when conducted on a large scale. Furthermore, as this approach primarily focuses on explicit, tangible research collaboration networks, it may overlook the identification of virtual, cross-organizational research teams, thus limiting its reference value.

Association rules are used to discover relationships and dependencies between sets of data items in large databases (Zou, 2009). In the context of identifying research teams, applying association rule algorithms involves utilizing the concept of frequent itemsets. By mining for frequent itemsets with support greater than a threshold SSS among researchers in a scientific institution, these sets are merged and cleaned, with the largest frequent itemsets being designated as research teams. Gao et al. (2014) applied frequent itemset extraction methods from association rules to analyze multi-word co-occurrences in the field of digital information transmission at China’s National Intellectual Property Administration, thus identifying relevant research hotspots. Similarly, Lv et al. (2016) employed the FP-Growth algorithm from association rules to excavate and identify core R&D teams at the Changchun Institute of Optics, Fine Mechanics and Physics of the Chinese Academy of Sciences, providing an in-depth interpretation of the seven identified core teams. They further verified their identification results through a questionnaire survey, confirming the method’s effectiveness.

The identification method based on association rules focuses on mining the largest frequent itemsets that represent the closest collaboration to recognize research teams, making it suitable for large-scale author collaboration data. However, this method does not emphasize the role of team leaders and fails to identify non-core researchers, which diminishes the reference value of the identification results.

Cluster analysis is a method that aggregates and classifies datasets based on similarities and differences under specific algorithmic rules, ensuring high similarity among individuals within the same group and significant heterogeneity between different groups (Wang, 2007; Wang and Jiao, 2013). This method divides samples into several clusters, allowing researchers to conduct further analysis based on the characteristics of these clusters in conjunction with other statistical methods, such as univariate analysis (Chen et al., 2023). The research team identification method based on hierarchical clustering involves two main steps: first, defining indicators that describe the similarity between nodes. Common calculation methods include node exclusivity, edge exclusivity, edge clustering coefficients, edge centrality, flow methods, resistance network methods, and random walk methods. Second, these similarity indicators are used to compute the similarity between pairs of nodes and extract communities. Community extraction methods can be broadly categorized into agglomerative and divisive approaches: agglomerative clustering works from the bottom up, while divisive clustering operates from the top down (Balakrishana and Deo, 2006).

Perianes-Rodríguez et al. (2010) utilized hierarchical clustering, using factor analysis of the raw data matrix and co-authorship analysis results to measure node similarity, thereby identifying research teams. Yu et al. (2011) analyzed collaborative situations of multidisciplinary teams based on author collaboration networks, employing centrality measures, component analysis, and hierarchical clustering methods to identify research teams. However, there are certain issues with the calculation methods for different similarity indicators within hierarchical clustering for research team identification. For instance, node exclusivity and edge exclusivity metrics struggle to effectively address core nodes in networks, while edge clustering coefficients perform poorly when dealing with triangular networks.

Social Network Analysis (SNA) is a quantitative analytical method developed by sociologists using mathematical techniques and graph theory to study social relationships and interactions between individuals or organizations (Oliveira and Gama, 2012). This method encompasses various tools and techniques aimed at revealing structural features of networks, such as centrality, clustering, and homogeneity. Community detection is a critical issue in SNA, focused on unsupervisedly partitioning individuals with social connections into tightly-knit groups while maximizing separation between these groups (Chunaev, 2020).

In related research on discovering research teams, SNA employs authorship and citation data to construct an overall research collaboration network, using centrality metrics to identify key researchers and closely collaborating teams. The results are then visualized, enhancing the accuracy and credibility of team identification through the integration of human cognitive judgment with data analysis (Tang, 2010).

Newman (2001) were pioneers in constructing co-authorship networks based on published paper data from scholars, proposing the Girvan-Newman algorithm to partition research teams according to these networks. Liu et al. (2011) applied SNA to co-authored papers in the field of genetic engineering from 2000 to 2009, using a combination of two-mode and snowball sampling methods to identify research teams. They explored factors influencing team performance from the perspectives of network characteristics and core authors.

Kang et al. (2019) and Gao and Li (2021) integrated the Latent Dirichlet Allocation (LDA) model and Subject-Action-Object (SAO) semantic analysis with SNA to uncover potential research partners and identify research teams. Hao et al. (2020) utilized SNA based on 14,913 academic papers published in Science from January 2000 to June 2018 to construct collaboration network models for scholar nodes and paper nodes, examining the relationship between centrality in research collaboration networks and academic influence through correlation and OLS regression analyses. Gao and Zhou (2021) emphasized the significant influence of authors’ institutions and order of authorship by introducing a directed co-authorship network model, replacing traditional undirected network analysis to more effectively identify and define the structure of core research teams.

Lv and Tan (2022) applied the Louvain algorithm for community detection in information science research, optimizing research team identification by analyzing the collaboration intensity among researchers in specific interdisciplinary fields, achieving efficient identification of research teams within a large-scale data context. Additionally, some scholars have divided literature data into time periods to identify research teams over different times, tracking the evolution of these teams. Hopcroft et al. (2004) employed cosine similarity with centroid clustering to identify “natural communities” within large-scale scientific paper networks, revealing network evolution characteristics and emerging community structures through observations over time.

The research team identification method based on social network analysis combines human cognitive abilities with computational power, enhancing the credibility and accuracy of the identification results. However, this method relies on intermediary centrality rankings to determine team leaders, which may lead to multiple high-ranking researchers within the same team, erroneously splitting a single team into multiple ones and affecting the validity of team identification. Furthermore, this approach faces significant computational complexity when dealing with large-scale and high-dimensional data, lacking the capability for automatic feature extraction and adaptive processing.

The four methods of research team identification discussed above each have their unique features and applicable scenarios. The questionnaire method directly collects data, suitable for studying physical teams, but is time-consuming and lacks focus on virtual teams. The association rule algorithm identifies teams through data mining of frequent item sets, making it suitable for large datasets, yet it overlooks team leaders and non-core members. The cluster analysis method categorizes data based on similarity, useful for extracting patterns from complex data, though the choice of similarity indicators can impact the results. Social network analysis reveals the relationships and interactions of research teams through structural features such as centrality and clustering, capable of handling large-scale data while providing visual results. This approach combines computational capabilities with human judgment, enhancing the accuracy and credibility of team identification. Given its comprehensive analysis of research collaboration networks, SNA is particularly effective at identifying core members and key relationships within complex and dynamically changing research environments. Thus, this method is selected as the primary approach for this study, enabling a more effective revelation of the true structures and collaboration patterns of research teams.

In light of this, this study employs social network analysis, primarily utilizing author collaboration network clustering to identify closely collaborating research groups. First, 33 leaders of research teams from universities participating in the “Open Bidding for Selecting the Best Candidates” initiative are randomly selected based on data from the “Open Bidding for Selecting the Best Candidates” policy text database. Detailed information about these leaders—such as their workplaces, areas of expertise, and educational backgrounds—is collected and matched from various sources, including Baidu Baike, the China National Knowledge Infrastructure (CNKI), and university official websites. Subsequently, based on this public information, the papers published by these leaders in core journals and CSSCI-indexed journals from 2000 to February 2024 are downloaded from the CNKI database. Using Gephi software, author collaboration network clustering is conducted, filtering research teams based on collaboration frequency, as shown in  Appendix. This methodology enables the construction of an author collaboration network and the compilation of a list of research teams, resulting in the identification of the research teams associated with these 33 university-based “Open Bidding for Selecting the Best Candidates” leaders, as shown in Table 2.

The table above lists relevant information about 33 leaders of research teams from universities participating in the “Open Bidding for Selecting the Best Candidates” initiative and their respective research groups. It is evident that these randomly selected leaders span a wide range of disciplines, from criminal law and international trade to the philosophy of science and technology, showcasing the breadth and diversity of research in higher education.

Regarding publication output, the data collected for these leaders covers their works published after 2000. Notably, most teams have published articles between 2000 and 2024, indicating a significant research history and stable output, with publication counts ranging from a minimum of 7 to a maximum of 235 articles. This suggests that there are variations in research productivity among the teams, yet all demonstrate a level of active engagement in research.

In terms of team characteristics, certain teams excel in specific fields; for instance, Team 12 has published 235 articles in the field of business management, reflecting their substantial expertise and influence in that area. Additionally, some teams stand out for their solo publications, such as Team 18, which has produced 117 single-authored articles, highlighting the independence and contributions of the team leader in their research endeavors.

The composition of these research teams shows an average membership of around five members, indicating a relative stability and sustainability in team size within university research settings. The close collaboration and complementary strengths among team members provide a solid foundation for in-depth research activities. Overall, the research teams led by these 33 “Open Bidding for Selecting the Best Candidates” leaders exhibit notable research capabilities and influence in their respective fields.

This study focuses on the 33 leaders of research teams involved in the “Open Bidding for Selecting the Best Candidates” initiative, employing a comprehensive evaluation framework based on four primary indicators: research output capability, continuity of research direction, interdisciplinary integration ability, and team collaboration intensity. Secondary indicators, including the publication volume of the leaders, will also be considered in assessing the performance of these research teams.

First, as illustrated in Table 3 and Figure 1, an in-depth analysis of the overall publication volume of the research team leaders reveals that, between 2000 and 2024, the 33 leaders examined published an average of approximately 69 academic papers each. This indicates that these prominent researchers have maintained a relatively stable research output over an extended period, serving as significant drivers of knowledge innovation and academic contribution within their respective fields. Among this group, the leader with the highest publication volume authored an impressive 235 papers.

Focusing on the period from 2018 to 2022, this study closely compares the publication trends of these leaders before and after they received their talent honors. The data analysis indicates that while the overall change in publication volume before and after the recognition is relatively limited, it is noteworthy that most leaders experienced a slight decline in their publication output post-honor compared to the years leading up to their selection. This phenomenon may reflect that, although receiving such titles significantly enhances their status and influence in academia, the quantitative measure of direct research output—namely, publication volume—may have been affected. Potential reasons for this decline could include a shift in research focus, increased administrative responsibilities, or a pursuit of higher quality in research outputs following their recognition.

As shown in Figure 2, this study calculates the proportion of papers authored as the first author or corresponding author by the leaders of the “Open Bidding for Selecting the Best Candidates” research teams published in top domestic journals relative to their total publication volume. The results reveal significant variability in this ratio, ranging from 0 to 0.62. This notable difference may be attributed to the varying number of top-tier journals across different academic fields and the specific research characteristics inherent to each discipline. For instance, fields such as mechanical design and control theory typically have fewer top-tier journals compared to areas like economics and management. Furthermore, many research outputs in engineering disciplines are highly practical and application-oriented, leading researchers to prioritize patent applications as a means of protecting technological achievements and promoting technology transfer, rather than solely relying on academic publications.

Among the 33 leaders included in the analysis, the average proportion of papers published in top journals reaches 0.24. This figure indicates that the “Open Bidding for Selecting the Best Candidates” research teams have achieved a certain level of effectiveness in producing high-quality research outcomes, though there remains substantial room for improvement in output quality. Notably, ten teams have a proportion of top-tier journal publications exceeding 40%, showcasing their high-level research capabilities and significant impact within their respective fields. Especially noteworthy is the fact that five teams have a publication proportion in top journals exceeding 50%, signifying that their research outputs have been recognized by peer experts for their innovative content, depth of knowledge, and academic novelty. This reflects the teams’ leading positions at the forefront of scientific research. Additionally, most of these publications in top journals are co-authored, indicating that the team leaders can effectively guide and motivate their members to engage in high-quality research activities, fostering knowledge sharing and collaborative thinking, which in turn accelerates the process of scientific discovery.

Through a comprehensive analysis of the keywords and abstracts of academic papers authored by the leaders of the “Open Bidding for Selecting the Best Candidates” research teams during the five years before and after they received their prestigious titles, this study systematically investigates whether their research directions have undergone significant changes. According to the results presented in Table 4, it is evident that, with the exception of the leader of Team 19, the research directions of the other 32 leaders exhibit a high degree of consistency before and after their selection for the titles, showing no notable changes. This phenomenon strongly underscores the remarkable stability and continuity of research directions within the “Open Bidding for Selecting the Best Candidates” research teams. This stability may be attributed to their sustained focus on in-depth exploration and theoretical expansion within specific fields, reflecting a solid academic foundation and meticulous work in their areas of expertise. Furthermore, it emphasizes the importance of long-term research planning and deep engagement in a field for fostering research innovation and building cohesive teams.

Based on an in-depth analysis of keywords and abstracts in the published papers of the research teams, this study evaluates the strength of interdisciplinary connections among various fields. Specifically, a research team is considered to exhibit strong interdisciplinary integration and cross-disciplinary research capabilities if it frequently includes keywords or themes that span multiple disciplines, with high co-occurrence rates. The results indicate that, on average, the 33 research team leaders examined span 1.7 academic fields in their work (as shown in Table 4). More specifically, 17 team leaders focus on a single discipline, while 10 teams cross two fields, demonstrating a certain trend toward interdisciplinary collaboration. Additionally, four teams engage in research across three disciplines, showcasing more pronounced interdisciplinary characteristics, and only two teams cover four fields, indicating a higher level of disciplinary integration. However, from an overall perspective, these 33 “leading” research teams appear relatively weak in their interdisciplinary capabilities, as most prefer to delve deeply into specific fields rather than broadly exploring multiple disciplines. Therefore, to enhance the breadth and depth of their research innovation, these teams should broaden their research perspectives and strengthen their interdisciplinary collaboration to better address the complex and evolving challenges in scientific research.

This study examines the collaboration intensity within research teams by analyzing the total number of papers published by team leaders and the number of co-authored papers. As illustrated in Figure 3, the average co-authorship rate among the 33 research teams is 0.66, indicating that more than half of the research outputs from the leaders are collaboratively produced with team members. This finding significantly reflects the prevalent phenomenon of close cooperation within the teams and further validates the active role of team leaders in fostering interaction, guidance, and mentorship among members.

Notably, eight research team leaders published only co-authored papers between 2000 and 2024, highlighting a highly integrated collaborative model within these teams. This not only demonstrates the strong academic synergy and cooperative spirit among team members but also has a positive impact on enhancing overall research effectiveness. Conversely, four research team leaders have a co-authorship ratio with their team members that is less than 15% of their total publications. This suggests a relatively lower level of collaboration within these teams, indicating that leaders may prefer to publish independently, which may hinder effective communication and cooperation among team members. Consequently, this could adversely affect the overall research quality and outcomes of the team.

This study analyzed a sample of 33 “Open Bidding for Selecting the Best Candidates” research team leaders randomly selected from a policy text database, all affiliated with universities. By collecting publication data from the CNKI database for these leaders between 2000 and February 2024 in core journals and CSSCI-indexed journals, the research utilized social network analysis methods to identify their research teams based on collaborative publishing patterns. An indicator system was constructed, encompassing four primary metrics—research output capability, continuity of research direction, interdisciplinary integration ability, and team collaboration intensity—along with various secondary indicators, including publication volume of the leaders, to comprehensively evaluate the research teams.

The identification of these research teams revealed that the 33 leaders spanned a wide range of disciplines, from criminal law and international trade to the philosophy of science and technology, showcasing the breadth and diversity of academic research in higher education.

In evaluating research performance, it was found that the 33 leaders had an average publication volume of about 69 papers from 2000 to 2024, indicating a relatively stable output over an extended period. Notably, all leaders received honors in talent titles between 2018 and 2022; although their publication volume fluctuated minimally, most exhibited a slight downward trend in output post-honor, suggesting a subtle influence of academic recognition on research pace and strategic adjustments. Additionally, the average proportion of papers published in top-tier journals was around 24%, with five teams achieving a proportion exceeding 50%. Most papers in these prestigious journals were co-authored, reflecting the leaders’ ability to guide and motivate team members in high-quality research activities, thus fostering knowledge sharing and intellectual collaboration.

Regarding continuity of research direction, a detailed analysis of the keywords and abstracts of academic papers from the leaders before and after receiving their talent titles revealed that, with a few exceptions, the vast majority maintained a high degree of consistency in their research focus, demonstrating stability and continuity. This finding underscores their long-term commitment to in-depth exploration within specific fields, showcasing strong academic focus and capability in deepening expertise.

In terms of interdisciplinary integration ability, the study found variations among the 33 teams based on the keywords and abstracts of their published papers; however, overall, their integration capabilities were relatively weak. Specifically, over half of the teams concentrated their research primarily within a single discipline or crossed only two fields, indicating a preference for disciplinary specificity rather than broad interdisciplinary engagement. While a few teams displayed strong interdisciplinary characteristics by encompassing three or even four fields, they constituted a minority. Therefore, this study suggests that these “Open Bidding for Selecting the Best Candidates” research teams from universities need to enhance their interdisciplinary integration capabilities by broadening their research perspectives and strengthening cross-disciplinary collaboration to better address complex and evolving research challenges, ultimately promoting the breadth and depth of scientific innovation.

In terms of team collaboration intensity, the data showed an average co-authorship rate of 0.66 across the 33 research teams, indicating that more than half of the research outputs were achieved through close collaboration among team members. Remarkably, eight team leaders published exclusively co-authored papers between 2000 and 2024, highlighting a high level of internal collaboration and knowledge sharing. This also indicates that the leaders play a crucial role in enhancing team cohesion and effectively guiding and nurturing their members.

In conclusion, the leaders of these “Open Bidding for Selecting the Best Candidates” research teams from universities possess strong foundational theoretical knowledge and innovative capabilities, exhibit significant continuity in their research directions, and often delve deeply into a specific field. However, their interdisciplinary integration capabilities are relatively weak, making them more suitable for foundational theoretical research projects. For key technological projects that are crucial in the “Open Bidding for Selecting the Best Candidates” system, it is essential to move beyond the conventional thinking of focusing solely on background, title, or network. Instead, the selection should be oriented towards professional capability, forming interdisciplinary teams with diverse academic backgrounds and fields of expertise.

Funding: This study was supported by the Program for Innovation Research at the Central University of Finance and Economics.

Figure A1