=Paper=
{{Paper
|id=Vol-3745/paper21
|storemode=property
|title=Relationship between Team Diversity and Innovation Performance in Interdisciplinary Research Teams within the Field of Artificial Intelligence: Decision Tree Analysis
|pdfUrl=https://ceur-ws.org/Vol-3745/paper21.pdf
|volume=Vol-3745
|authors=Junwan Liu,Chenchen Huang,Shuo Xu
|dblpUrl=https://dblp.org/rec/conf/eeke/LiuHX24
}}
==Relationship between Team Diversity and Innovation Performance in Interdisciplinary Research Teams within the Field of Artificial Intelligence: Decision Tree Analysis==
https://ceur-ws.org/Vol-3745/paper21.pdf
Relationship between Team Diversity and Innovation Performance in
Interdisciplinary Research Teams within the Field of Artificial
Intelligence: Decision Tree Analysis
Junwan Liu1,*, Chenchen Huang1, and Shuo Xu1
1 College of Economics and Management, Beijing University of Technology, Beijing, 100124, China
Abstract
Interdisciplinary research teams are crucial in solving complex problems by providing creative solutions that single-
discipline teams cannot achieve. Previous studies have primarily focused on the linear relationship between independent
variables and team innovation performance, neglecting the non-linear aspect. To address this gap, this paper examines the
non-linear relationship between diverse factors and the innovation performance of interdisciplinary research teams in
artificial intelligence. By utilizing the Classification and Regression Tree (CART) model, the study reveals that activity
diversity and interdisciplinary research team innovation performance exhibit a U-shaped relationship in terms of “novelty”
innovation performance. Furthermore, this relationship is influenced by research interest diversity. Specifically, low
research interest diversity leads to low innovation performance as activity diversity increases. Meanwhile, research interest
diversity emerges as the most critical factor impacting innovation performance. The importance of member diversity,
institutional diversity, and activity diversity on innovation performance should not be ignored. Through decision tree
analysis, this paper extends research on the multifactor combination, complex nonlinear relationships, and multipath
influence mechanism of team diversity on interdisciplinary research teams’ innovation performance.
Keywords
Interdisciplinary research team, Team diversity, Innovation performance, Classification and regression tree (CART) model
diversity and task-related diversity[6]. Demographic diversity
1. Introduction* encompasses variations in team members’ demographic attributes,
Growing globalization and intense market competition have such as age, gender, and institutional backgrounds[7]. Task-
transformed the scientific model and increased the number of related diversity pertains to the diverse qualities that team
specialized research teams. In order to enhance the efficiency and members bring to their academic or professional pursuits,
quality of scientific research, more research teams are including workplace functions, knowledge, and education[8].
transitioning from single teams to diversified teams[1]. Creating successful teams with demographic and task-related
Interdisciplinary research has become a necessary choice in this diversity is not a straightforward process of simply combining
context[2], allowing teams to draw upon a wide range of individuals from different disciplines. Horwitz et al.[6] discovered
disciplines and expertise to address complex scientific questions. that while demographic diversity did not significantly impact team
The diversity of team members from various disciplines and performance, task-related diversity positively influenced it.
backgrounds contributes to greater knowledge and innovative Diverse teams struggle with issues such as gender differences,
results[3]. Thus, effective interdisciplinary collaboration is crucial team conflict, and collaboration[9]. Given these contradictory
for achieving scientific and innovative breakthroughs[4]. findings, our focus is on investigating the impact of both
Previous research highlights the significance of diversity as a demographic diversity and task-related diversity on the innovation
crucial factor influencing the success of interdisciplinary research performance of interdisciplinary research teams, while analyzing
teams[5]. Diversity can be broadly categorized as demographic the varying importance of different diversity factors. Existing
studies have primarily focused on exploring the linear relationship
between independent variables and team innovation performance,
Joint Workshop of the 5th Extraction and Evaluation of Knowledge overlooking the nonlinear aspect. The nonlinear relationship
Entities from Scientific Documents and the 4th AI + Informetrics (EEKE- between these characteristics and team innovation performance,
AII2024), April 23~24, 2024, Changchun, China and Online
* Corresponding author. especially in the context of demographic diversity and task-related
EMAIL: liujunwan@bjut.edu.cn (Junwan Liu) diversity in interdisciplinary research teams, remains unclear.
© Copyright 2024 for this paper by its authors. Use permitted under Creative This paper aims to investigate the impact and decision-making
Commons License Attribution 4.0 International (CC BY 4.0).
mechanisms of team diversity on the innovation performance of
interdisciplinary research teams. Firstly, team innovation
CEUR
ceur-ws.org
Workshop ISSN 1613-0073
Proceedings
134
performance is divided into novelty and impact[10]. Second, we This paper focuses on empirical research in the field of
will investigate the influence of team diversity on the artificial intelligence (AI). The dataset is derived from the
interdisciplinary research team’s innovation performance in terms information of the most influential scholar award winners on the
of both the demographic diversity and task-related diversity of AMiner website (https://www.aminer.cn/ai2000) 2023 AI 2000
team members. From a social categorization perspective, we annual list. There are three reasons for selecting these scholars as
assume that gender diversity, national diversity, and institutional the subjects of the study. Firstly, AI research is inherently
diversity are included in the demographic diversity in this context. interdisciplinary[11]. Second, since its launch in 2006, the
Meanwhile, relying on the informational decision-making AMiner platform has already been used by many researchers[12].
perspective, we hypothesize that task-related diversity includes Third, since 2017, the AMiner platform has been publishing the
sociability diversity, activity diversity, research interest diversity, annual AI 2000 most influential scholar list. The purpose of this
and member diversity. Specifically, we address the following list is to annually rank the 2000 scholars who are expected to be
research questions in this paper: RQ1: What is the complex highly cited in the field of AI over the next ten years (2020-2029).
relationship structure among demographic diversity, task-related
diversity, and the innovation performance of interdisciplinary 2.2 Team Recognition
research teams? RQ2: What combinations of characteristics
Firstly, we obtain the dataset from the AMiner website, which
promote high levels of team innovation performance? RQ3: What
covers information on the most influential scholars in the 2023 AI
diversity characteristics should researchers focus on to enhance
2000 annual list. The dataset consists of 195 selected scholars and
the innovation performance of interdisciplinary research teams?
their collaborators, with five of these scholars receiving awards in
two or more subfields. The public information of the selected
2. Data and methods scholars is obtained from their personal websites and academic
This paper investigates the impact of team diversity on social networking platforms. The papers of selected scholars are
interdisciplinary research teams’ innovation performance. The downloaded from the Web of Science database. Finally, 25,285
basic process is shown in Figure 1. Firstly, raw data is processed papers are submitted by 195 selected scholars.
to form authors’ collaborative relationship data and measure each Second, we conduct community detection on the author co-
author’s collaborative tie strength. Second, stable collaborative occurrence network for each selected scholar using the Louvain
relationships are identified using a pre-set threshold (super tie). algorithm. This algorithm divides the nodes in the network into
Then, members of interdisciplinary research teams are identified, different communities based on modularity metrics, which assess
and the diversity index of each team is measured. Again, team the collaborative relationships between the nodes. The algorithm
innovation performance is divided into novelty and impact to be identifies strong connections within the same community and
measured. Finally, by using team diversity as the conditional sparser connections between different communities. In the
attribute and innovation performance as the decision attribute, the network, each co-author is represented as a node, and edges
impact of team diversity on the interdisciplinary research team’s represent collaborations between selected scholars and co-authors
innovation performance is explored using the CART model. who have published papers together. Figure 2 shows the author
Data Processing co-occurrence network for Silver, D selected scholars. Different
Paper publication Collaborative tie colors in the figure represent various communities determined by
data strength
Collaborative
Strong tie
Interdisciplinary Research modularity, with the community to which Silver and other
Author Networks Teams
collaborative Super tie
selected scholars belong shown in purple.
relationship
Team
Diversity
Innovation
gender institutional national sociability performance
diversity diversity diversity diversity
activity member
research interest diversity Novelty Impact
diversity diversity
CART Data Training
Decision tree
classification result x1 CART algorithm to obtain decision rules
Decision tree
classification result x2 What kind of diversity has an impact
on the interdisciplinary research
teams' innovation performance?
······ ······
Figure 2: Collaboration Network of Silver, D scholar
Figure 1: Research Framework
Next, we calculate the collaborative tie strength among the
nodes in each community to filter out the core collaborators of
2.1 Data collection each selected scholar. Collaborative tie strength, also known as a
“super tie”, has been extensively studied in scientific collaborative
135
networks[13]. It represents a long-term and stable collaboration, 35
similar to life partners, characterized by high intensity, close ties, 30
and long durations[14]. To identify the core collaborators among 25
Team Number
the 195 selected scholars, we calculate the super tie for each 20
community. Specifically, when a member’s collaborative tie
15
strength exceeds his or her community’s super tie threshold, that
10
member is referred to as a super tie collaborator and core team
member. Equations 1 and 2 demonstrate the formula for 5
calculating the super tie[14]. Additionally, we employ the 0
2 3 4 5 6 7 8 9 10 11 12 13 14 17 18 22 23 28 33 39
Anderson-Darling test to examine the distribution of collaborative
Team Size
intensity 𝐾𝑖𝑗 /𝐾𝑖 among the members. Our analysis indicates that
the statistical distribution 𝑃(𝐾𝑖𝑗 ) of all members’ collaborative Figure 3: Size Distribution of Interdisciplinary Research Teams
intensity conforms to an exponential distribution, with the average Table 1
collaborative strength of members being 2.83. Distribution of Members from Different Disciplinary Fields
〈𝐾𝑖 〉 = 𝑆𝑖−1 ∑𝑆𝑗=1
𝑖
𝐾𝑖𝑗 (1) Discipline Percentage of members
𝐾𝑖𝑐 = (〈𝐾𝑖 〉 − 1) ln 𝑆𝑖 (2) Computer and information science 71.18%
Where the collaborative tie strength 𝐾𝑖𝑗 is defined as the Electrical engineering, electronic
21.80%
engineering, information engineering
cumulative number of papers co-authored by the selected scholar Environmental engineering 2.03%
in community 𝑖 and scholar 𝑗 over the time between their first and Nano-technology 1.52%
last paper. 𝑆𝑖 represents the number of different co-authors of Physical science 1.09%
selected scholars in the community 𝑖. 〈𝐾𝑖 〉 represents the average Health science 0.94%
collaborative tie strength 𝐾𝑖𝑗 . Each scholar 𝑗 with 𝐾𝑖𝑗 > 𝐾𝑖𝑐 is Clinical medicine 0.51%
Mathematics 0.36%
labelled as a super tie collaborator of community 𝑖.
Materials engineering 0.22%
Eventually, we identify 195 research teams and their 1,217 Medical engineering 0.22%
core members. Figure 3 shows the distribution of team sizes for Basic medicine 0.14%
195 teams. The largest team size is 57 members, and 165 teams
are smaller than 10 members, which represents 85% of all teams. 2.3 Variables
Previous research defines interdisciplinary teams as groups of
scientists from different disciplines who collaborate to address 2.3.1 Dependent variables
complex problems[15]. To verify the interdisciplinarity of these We calculate the degree of team novelty using the novelty index
teams, we utilize a method that maps member affiliations to proposed by Lee et al.[10]. This index measures the novelty of a
disciplinary classifications[16] to more accurately determine the team’s paper based on the rarity of prior citation pairs. The
disciplinary backgrounds of the members. Specifically, we extract
calculation involves two steps.
secondary institutions from each member’s address, retain the
Complete the first step of the operation on the paper level. (1)
disciplinary terms in the secondary institution names, and match
List all paired reference combinations for each paper. (2) Record
these terms to the discipline field in the OECD classification the corresponding journal pairs. (3) Aggregate the pairs of journal
scheme. In this way, each member’s institution can be precisely
combinations published from year t-2 to year t as 𝑈𝑡 set. The time
matched to his or her research discipline. The results indicate that
window from year t-2 to year t is chosen to ensure data robustness.
165 teams have members from two different disciplinary
We then calculate the commonness value using Equation 3[10]. (4)
backgrounds, 27 teams have members from three different
This equation assigns each paper a range of commonness values.
disciplinary backgrounds, and 3 teams have members from four
The commonness values of each paper are ranked, and the 10th
different disciplinary backgrounds, thus reinforcing that the 195
percentile is taken as the commonness value of the paper. Using
teams in this study are interdisciplinary research teams. Table 1
the 10th percentile instead of the minimum value helps reduce
demonstrates the distribution of members from different
noise and increase the reliability of the measure. (5) The
disciplinary fields. Specifically, 71.18% of the members are from
commonness value is transformed using a natural logarithm to
the field of computer and information science, 21.80% are from
obtain an approximately normally distributed variable. The final
the fields of electrical engineering, electronic engineering, and
novelty value for that paper is obtained by adding a negative sign.
information engineering, while other fields encompass 𝑜𝑏𝑠𝑒𝑟𝑣𝑒𝑑 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑝𝑎𝑖𝑟𝑠𝑖𝑗𝑡 𝑁𝑖𝑗𝑡
environmental engineering, nanotechnology, and physical Commonness𝑖𝑗𝑡 = = 𝑁𝑖𝑡 𝑁𝑗𝑡 =
𝑒𝑥𝑝𝑒𝑐𝑡𝑒𝑑 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑝𝑎𝑖𝑟𝑠𝑖𝑗𝑡 × ×𝑁𝑡
𝑁𝑡 𝑁𝑡
sciences, among others. To explore the factors influencing the 𝑁𝑖𝑗𝑡 ×𝑁𝑡
interdisciplinary research teams’ innovation performance, we (3)
𝑁𝑖𝑡 ×𝑁𝑗𝑡
download each team member’s papers from the Web of Science Where 𝑁𝑖𝑗𝑡 is the number of occurrences of journal pairs (i, j)
database and collect 91,025 papers from all teams. in 𝑈𝑡 set. 𝑁𝑖𝑡 is the number of journal pairs in 𝑈𝑡 set that contain
136
journal 𝑖. 𝑁𝑗𝑡 is the number of journal pairs in 𝑈𝑡 set that contain To assess the diversity and differences in a team’s overall
journal 𝑗, and 𝑁𝑡 is the number of all journal pairs in 𝑈𝑡 set. sociability and activity, we utilize Equation 4 to calculate the
Calculating novelty at the team level is the next stage. The diversity of these two evaluation indicators.
number of paper publications by each team is counted and divided The research interest diversity of scholar is based on the
by the team size to calculate the team’s novelty value. breadth of the field of interest. Using the topic model, we identify
We measure a team’s impact using forward citations[10]. each scholar’s field of study and assign their papers to relevant
High-impact papers are defined as those in the top 1% of citation topics. The 𝑃𝐴 (t) topic distribution is obtained by Equation 7, and
distribution. This definition follows Uzzi et al.[17] and considers the research interest diversity is defined as the threshold of the
the use of short citation time windows can lead to the incorrect distribution of the 𝑃𝐴 (𝑡), which is calculated by Equation 8.
#𝑝𝑎𝑝𝑒𝑟𝑠 𝑜𝑓 𝐴 𝑏𝑒𝑙𝑜𝑛𝑔 𝑡𝑜 𝑡𝑜𝑝𝑖𝑐 𝑡
identification of highly cited papers[18]. First, the process of 𝑃𝐴 (t) = (7)
#𝑎𝑙𝑙 𝑝𝑎𝑝𝑒𝑟 𝑜𝑓 𝐴
identifying high-impact papers is completed at the paper level. (1)
Rank all papers from highest to lowest citation count. (2) By using research interest diversity(A) = − ∑ 𝑃𝐴 (𝑡)𝑙𝑜𝑔𝑃𝐴 (𝑡)
a five-year moving window, we define papers in the top 1% of the 𝑡∈𝑎𝑙𝑙 𝑡𝑜𝑝𝑖𝑐 𝑜𝑓 𝐴
rankings from year t-5 to year t as high-impact papers. (3) We use (8)
a dummy variable to indicate whether each paper is a high-impact Member diversity refers to the extent of diversity in
publication, assigning a value of 1 if it is and 0 otherwise. Next, collaborative relationships among team members. A low diversity
the number of high-impact papers per team is counted and divided indicates frequent collaboration with the same co-authors, while a
by the team size to finally obtain the team’s impact value. high diversity reflects collaboration with diverse co-authors, as
calculated by Equation 9[23].
𝑟𝑒𝑙𝑎𝑡𝑖𝑜𝑛𝑖
2.3.2 Independent variables member diversity = (9)
𝑐𝑜𝑎𝑢𝑡ℎ𝑜𝑟𝑖 (𝑛𝑜𝑛−𝑑𝑢𝑝𝑙𝑖𝑐𝑎𝑡𝑒)
Gender diversity refers to the subjective or objective similarities Where 𝑟𝑒𝑙𝑎𝑡𝑖𝑜𝑛𝑖 refers to the number of collaborations in
and differences between team members in terms of gender[19]. published journal articles, and 𝑐𝑜𝑎𝑢𝑡ℎ𝑜𝑟𝑖 (𝑛𝑜𝑛 − 𝑑𝑢𝑝𝑙𝑖𝑐𝑎𝑡𝑒)
National diversity is defined as having team members from represents the total number of non-replicating co-authors. Table 2
different national backgrounds, which introduces sociological is the examples of co-authors’ member diversity of member A.
categorization and the potential for diverse cognitive Table 2
perspectives[20]. Institutional diversity refers to the presence of a Examples of Co-authors’ Member Diversity of Member A
variety of members in different institutions[21]. All of the above Article 1 Article 2
demographic diversity indicators mentioned above are measured Author A, B, C, D A, C, D
using the Simpson index, which is calculated using Equation 4. Relation A-B, A-C, A-D A-C, A-D
𝑛
𝐻 = 1 − ∑𝑖=1 𝑃𝑖2 (4) Coauthor(non-duplicate) A, B, C, D
Member diversity 5/4=1.25
Where 𝑛 is the total number of categories, 𝑃𝑖 is the percentage
of members of the group 𝑖. The higher the 𝐻 value, the greater is
the value of the diversity. 2.4 Data Characteristics and Analysis
Using the AMiner platform, we algorithmically obtain data on Table 3 provides the results of the exploratory analysis, including
sociability, activity, and research interest diversity indices to averages, mediums, minimum, and maximum values for
assess the academic proficiency of members. Definitions and calculated indicators. Figure 4 and Figure 5 provide correlation
formulas for these indicators are provided[22]. plots between diversity indicators and interdisciplinary research
The sociability index is derived from considering both the teams’ innovation performance. There is no strong correlation
number of scholars’ collaborators and their collaborative papers, between population diversity, task-related diversity and
as shown in Equation 5. innovation performance.
𝑠𝑜𝑐𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑦(𝐴) = 1 + ∑𝑒𝑎𝑐ℎ 𝑐𝑜𝑎𝑢𝑡ℎ𝑜𝑟(𝑐) 𝑙 𝑛(#𝑐𝑜𝑝𝑎𝑝𝑒𝑟𝑐 ) (5)
Where the #𝑐𝑜𝑝𝑎𝑝𝑒𝑟𝑐 is the number of papers co-authored
between scholar and co-authors.
The activity index measures a scholar’s frequency and number
of recent publications, along with the significance of each paper,
as shown in Equation 6.
activity(A) = ∑𝑒𝑎𝑐ℎ 𝑦𝑒𝑎𝑟(𝑛)𝑖𝑛 𝑟𝑒𝑐𝑒𝑛𝑡 𝑁 𝑦𝑒𝑎𝑟𝑠 IS(𝐺𝑛 ) × weight(n)
(6)
Where, in 𝑛 years (𝑛 belongs to near 𝑁 years), 𝐺𝑛 is a group of
papers published by scholars in n years, weight(n) = 𝛼 this year−n ,
and the following principles are applied to the values of 𝑛 and 𝛼: Figure 4: Correlation between Diversity Indicators and Team’s
if the current month is in the first half of the year (month -0.40 > -0.31 - Low 29.00% 61.00%
handling complex relationships. These models rely on the least
Novelty - - - - > -0.40 <= -0.31 - High 12.00% 79.00%
squares method, which requires rigorous hypothesis testing and
- - - - <= -0.40 - - High 14.00% 66.00%
variable control[26]. In the context of understanding how
- <= -0.72 - - - <= 0.47 <= -0.08 Low 9.00% 78.00%
diversity impacts the teams’ innovation performance, it is crucial
- > -0.72 - - - <= 0.47 <= -0.08 High 31.00% 60.00%
to consider multiple independent variables and potential
interactions among them. Incorrect selection of control variables - - - - - > 0.47 <= -0.08 Low 7.00% 86.00%
Impact
or omission of important variables may lead to biased results in - - - - - <= -0.29 > -0.08 High 8.00% 93.00%
regression analysis. In contrast, the CART model offers a more - - - - - (-0.29,0.08] > -0.08 Low 5.00% 70.00%
flexible and robust solution[27]. It constructs a decision tree using - - - - - (0.08,0.47] > -0.08 High 19.00% 74.00%
recursive binary splitting, dividing data subsets into smaller Firstly, in the “novelty”, interdisciplinary research teams have
subsets[24]. This nonparametric approach eliminates the need for a higher proportion of high innovation performance ratings. In the
138
“impact”, there is a higher proportion of interdisciplinary teams to prioritize the diversity of institutions represented within the
with a low innovation performance rating than in the “novelty” teams.
innovation performance. Secondly, activity diversity and member
diversity split as root nodes, which are key factors affecting the
novelty and impact of a team, respectively. Third, the confidence
coefficients for most of the decision rules are above 60%,
indicating that the weight of the sample size supporting the ≤ >
current decision rule in the leaf node’s sample size is 60% or more.
This suggests that the results are highly interpretable.
Figure 6 shows that there are a total of three rules to determine
whether a team has high or low innovation performance. The
CART model divides data into approximately two branches based
on whether the most important feature “activity diversity” is less
than or equal to -0.40. The results show that with lower activity
diversity, team members can focus more on their original thinking,
enhancing team innovation performance without the need to be
concerned about publication frequency and quantity. When
activity diversity is higher, an increase in research interest
diversity contributes to teams achieving high levels of innovation
Figure 7: Decision Tree for Team’s Impact
performance. Horizontal comparison reveals that activity diversity
is a crucial factor influencing innovation performance.
Interdisciplinary research teams with low activity diversity are not
3.2 Feature Importance Analysis
influenced by research interest diversity, whereas interdisciplinary The model’s final characteristic importance for explanatory
research teams with high activity diversity are impacted by variables is presented in Figure 8. Among the factors affecting a
research interest diversity. team’s novelty, research interest diversity has the highest
characteristic importance of 0.97, while activity diversity has a
characteristic importance of 0.03. Among the factors affecting a
team’s impact, research interest diversity has the highest
characteristic importance of 0.48. Member diversity and
institutional diversity closely followed with 0.31 and 0.22,
respectively. This result shows that research interest diversity is
≤ >
most strongly associated with interdisciplinary research teams'
innovation performance. This shows that team members' diverse
expertise backgrounds enable knowledge integration and
reconfiguration, which are crucial elements in innovation
performance[21].
0.03 Novelty
AD 0 Impact
MD 0
0.31
SD 0
0
0.97
Figure 6: Decision Tree for Team’s Novelty RID 0.48
Figure 7 shows six rules to determine whether a team has high ND 0
0
or low innovation performance. The CART model divides the data 0
ID 0.22
into two branches based on whether the most important feature 0
“member diversity” is less than or equal to -0.08. The results GD 0
reveal that interdisciplinary research teams with member diversity
more than -0.08 need to control research interest diversity to 0.0 0.2 0.4 0.6 0.8 1.0
achieve high innovation performance. For interdisciplinary Figure 8: Characteristic Importance of Explanatory Variables
research teams with member diversity smaller than -0.08,
institutional diversity is a key influence on innovation 4. Conclusion
performance. Horizontal comparison shows that member diversity
is a key factor influencing innovation performance. This paper finds a U-shaped relationship between activity
Interdisciplinary research teams with low member diversity tend diversity and team innovation performance in “novelty”
innovation performance. However, this relationship is impacted
by research interests diversity. Specifically, interdisciplinary
139
teams with low activity diversity are able to improve their [4] Salazar, M., Lant, T. (2018). Facilitating innovation in interdisciplinary teams:
The role of leaders and integrative communication. Informing Science, 21,
innovation performance independently of research interest 157.
diversity. In contrast, low research interest diversity leads to low [5] Benoliel, P., Somech, A. (2015). The role of leader boundary activities in
enhancing interdisciplinary team effectiveness. Small Group Research, 46(1),
innovation performance when activity diversity increases. In 83-124.
terms of “impact” innovation performance, increasing member [6] Horwitz, S. K., Horwitz, I. B. (2007). The effects of team diversity on team
diversity and managing the range of research interests can be outcomes: A meta-analytic review of team demography. Journal of
management, 33(6), 987-1015.
beneficial. In addition, interdisciplinary research teams with low [7] Lee, D. S., Lee, K. C., Seo, Y. W. (2015). An analysis of shared leadership,
member diversity need to focus on the institutional diversity of diversity, and team creativity in an e-learning environment. Computers in
Human Behavior, 42, 47-56.
team members, as institutional diversity has a positive impact on [8] Jackson, S. E., Joshi, A., Erhardt, N. L. (2003). Recent research on team and
the team’s effectiveness. organizational diversity: SWOT analysis and implications. Journal of
In the evaluation of various factors, research interest diversity Management, 29(6), 801-830.
[9] Abramo, G., D’Angelo, C. A., Di Costa, F. (2018). The effects of gender, age
emerges as the most significant determinant of innovation and academic rank on research diversification. Scientometrics, 114(2), 373-
performance in interdisciplinary research teams. This implies a 387.
[10] Lee, Y. N., Walsh, J. P., Wang, J. (2015). Creativity in scientific teams:
close association between research interest diversity and the Unpacking novelty and impact. Research policy, 44(3), 684-697.
team’s ability to innovate. Researchers with distinct research [11] Zhuang, Y., Cai, M., Li, X., Luo, X., Yang, Q., Wu, F. (2020). The next
breakthroughs of artificial intelligence: The interdisciplinary nature of AI.
themes bring a diverse range of knowledge and contribute to the Engineering, 6(3), 245.
reconfiguration of knowledge by identifying and integrating [12] Wu, J., Ou, G., Liu, X., Dong, K. (2022). How does academic education
insights from different fields[30]. background affect top researchers’ performance? Evidence from the field of
artificial intelligence. Journal of Informetrics, 16(2), 101292.
Managers should consider research team diversity when [13] Li, Y., Li, N., Guo, J., Li, J., Harris, T. B. (2018). A network view of advice-
developing it. To create a healthy innovation environment, they giving and individual creativity in teams: A brokerage-driven, socially
perpetuated phenomenon. Academy of Management Journal, 61(6), 2210-
should pay attention to the heterogeneity of different 2229.
organizational and disciplinary backgrounds to which team [14] Petersen, A. M. (2015). Quantifying the impact of weak, strong, and super
members belong and strive to optimize the level of knowledge ties in scientific careers. Proceedings of the National Academy of Sciences,
112(34), E4671-E4680.
diversity in the team. Furthermore, managers should encourage [15] Wagner, C. S., Roessner, J. D., Bobb, K., Klein, J. T., Boyack, K. W., Keyton,
and promote activity diversity. Meanwhile, the costs of too much J., Rafols, I., Börner, K. (2011). Approaches to understanding and measuring
interdisciplinary scientific research (IDR): A review of the literature. Journal
research interest diversity need to be noted to avoid the of informetrics, 5(1), 14-26.
phenomenon of “too much of a good thing”. This is because [16] Liu, X., Bu, Y., Li, M., Li, J. (2024). Monodisciplinary collaboration disrupts
knowledge diversity among collaborative members in different science more than multidisciplinary collaboration. Journal of the Association
for Information Science and Technology, 75(1), 59-78.
research fields is often considered a double-edged sword. As [17] Uzzi, B., Mukherjee, S., Stringer, M., Jones, B. (2013). Atypical
Wang et al have found, increasing knowledge diversity leads to a combinations and scientific impact. Science, 342(6157), 468-472.
[18] Wang, J. (2014). Unpacking the Matthew effect in citations. Journal of
decline in social influence after a certain peak[21]. Informetrics, 8(2), 329-339.
Our study has limitations. Firstly, it focuses solely on the [19] Van Knippenberg, D., Schippers, M. C. (2007). Work group diversity. Annu.
Rev. Psychol., 58, 515-541.
AMiner platform in AI, limiting its scope and generalizability. [20] Ayub, N., Jehn, K. A. (2006). National diversity and conflict in multinational
Future studies should broaden the research fields. Secondly, the workgroups: The moderating effect of nationalism. International Journal of
sample of 195 interdisciplinary teams may not fully reflect AI Conflict Management, 17(3), 181-202.
[21] Wang, G., Gan, Y., Yang, H. (2022). The inverted U-shaped relationship
team diversity and complexity, potentially suffering from between knowledge diversity of researchers and societal impact. Scientific
sampling error. A more representative sample is needed. Thirdly, Reports, 12(1), 18585.
[22] Zhuo, L. I. N., Haohai, H. U. A. N. G. (2022). Top Experts Identification and
while we focused on diversity within teams, future research could Evaluation of International Cooperation on Artificial Intelligence in China.
explore diversity in other research team activities. Lastly, our Journal of Library and Information Science in Agriculture, 2022,34(1): 86-95.
team recognition method overlooks member turnover dynamics. [23] Liao, C. H. (2011). How to improve research quality? Examining the impacts
of collaboration intensity and member diversity in collaboration networks.
Future studies should introduce a dynamic analysis of member Scientometrics, 86(3), 747-761.
flow for more accurate core member identification. [24] Breiman, L. (1984). Classification and regression trees (1st ed.). Routledge.
[25] Aboseif, E., Hanna, A. S. (2023). Defining the Success Status of Construction
Projects Based on Quantitative Performance Metrics Thresholds. Journal of
ACKNOWLEDGMENTS Management in Engineering, 39(2), 04022073.
[26] Shmueli, G. (2010). To explain or to predict? Statistical Science, 25(3), 289-
This work was supported partially by the National Natural 310.
[27] Klusowski, J. M., Tian, P. M. (2024). Large scale prediction with decision
Science Foundation of China (Grant No. 72174016). Our gratitude trees. Journal of the American Statistical Association, 119(545), 525-537.
also goes to the anonymous reviewers and the editor for their [28] Pahmi, S., Saepudin, S., Maesarah, N., Solehudin, U. I. (2018).
valuable comments. Implementation of CART (classification and regression trees) algorithm for
determining factors affecting employee performance. In 2018 International
Conference on Computing, Engineering, and Design (ICCED) (pp. 57-62).
REFERENCES IEEE.
[29] Chen, W., Xie, X., Wang, J., Pradhan, B., Hong, H., Bui, D. T., Duan, Z., Ma,
[1] Dong, Y., Ma, H., Tang, J., Wang, K. (2018). Collaboration diversity and J. (2017). A comparative study of logistic model tree, random forest, and
scientific impact. arXiv preprint arXiv,1806.03694. classification and regression tree models for spatial prediction of landslide
[2] Van Noorden, R. (2015). Interdisciplinary research by the numbers. Nature, susceptibility. Catena, 151, 147-160.
525(7569), 306-307. [30] Zhang, L., Li, X. (2016). How to reduce the negative impacts of knowledge
[3] Taylor, A., Greve, H. R. (2006). Superman or the fantastic four? Knowledge heterogeneity in engineering design team: Exploring the role of knowledge
combination and experience in innovative teams. Academy of management reuse. International Journal of Project Management, 34(7), 1138-1149.
journal, 49(4), 723-740.
140