Team Assembly

1
Team Assembly

• Zeina Atrash, Yun Huang, Drew Margolin, Katya
  Ognyanova, Noshir Contractor

2
Overview

• General Research Goals Rationale
• Defining Teams
• Team Composition and Relations
• Operationalizing Team Outcomes
• Team Influence on Outcome -- Preliminary Analyses
• Future Directions

3
Research Goals and Rationale

• How do the mechanisms that motivate team assembly
  lead to different outcomes of team projects?
•  
• Sample Research Questions
• What are the theoretical mechanisms leading to
  the assembly of teams?
• How do effective potential teammates find each
  other?
• What subset of these mechanisms lead to the
  assembly of the most effective teams?
• How does previous experience with other team
  members effect outcomes?
• Methodological Rationale
• A space of potential teammates -- NanoHub users
• A set of actual teams -- All NanoHub projects and
  subsequent publications
• A set of team outcomes -- usage and evaluation of
  projects, citation in publications

4
(No Transcript)
5
Defining and Identifying Teams

• We define teams as the collaborators or
  co-producers of a NanoHub related project.  
• NanoHub projects include
• NanoHub resources, (e.g. learning modules,
  courses)
• NanoHub simulation tools
• nanotechnology publications authored by NanoHub
  users.
• Any project that has more than a single author is
  produced by a "team."  
• The authors of that project are the team
• Teams can be unique to projects (i.e. they may or
  may not repeat)

6
Descriptive Statistics -- Team Composition

• In the current dataset, we identify
• 130 Team instances
• 129 Tools (112 teams, 92 unique)
• 45 Publications (18 teams)
• 170 Individual contributors to tools 46 to
  publications 
• Tools  Average team size N 3.1, SD1.55
• Publications Average team size N 2.22, SD2.10
  
• Team members (tools)
• From 32 different institutions (72 from Purdue)
• From 10 different countries (153 from the US)
• 144 male, 13 female, 13 unknown

•  

7
Team Composition -- Tools 
8
Contributors by Country
9
Defining and Measuring Team Outcomes

• Team outcomes can be measured by the success of
  the team's project
• Measures using nanoHub data and logs
• Average rating
• Variability of rating
• of times rated
• of jobs (for tools)
• of times accessed (Apache log data)
• of times tagged
• of subsequent collaborations emerging from this
  team
• Measures using external data
• of times cited in publications (nanoHUB, WoS,
  Google scholar)
• and impact of subsequent collaborations
  emerging from this team

10
Descriptive Statistics - Outcomes

• Tools (total 129)
• Average of times rated N1.16, SD1.23 (0 for
  71 tools)
• Average rating (1-10) N4.54, SD1.20  (if
  rated at least once)
• Average of users N397.66, SD572.83  
• Average of jobs   N6772.36, SD15233.11
•  
• Publications (total 45)
• Average of times rated N1.16, SD1.23 (0 for
  35 pubs)
• Average rating (1-10)N4.2, SD0.82 (if rated
  at least once)
• Average of users N394.84, SD441.34
•  

11
Descriptive Statistics - Outcomes Distribution
12
Potential Assembly Mechanisms

• NanoHub has 784 contributors.  What factors
  influence the choice of a few select teammates
  for particular projects?
•  
• Control measures --
• Association with Purdue
• Reputation or rank
• U.S. based / International
• Team size -- Do teams gain efficiency from large
  resource bases or do these lead to complexity and
  confusion?
• Homophily / diversity - How do teams balance the
  need for shared attributes and understandings
  with the need for complementary skills and
  perspectives?
• Network position -- Do some potential
  collaborators become attractive because of their
  connections to others?
• History - Do teams emphasize familiarity or
  novelty?

13
Measuring Team Composition

•  Homophily and Diversity Scales can be created
  that measure the degree to which members of a
  given team are similar or different in terms of
• Organization (university) 
• Position (professor, student, etc.)
• Training and expertise (department)
• Research interests (tags, subject categories,
  keywords)
• Geographic distance (city of origin,
  International vs. domestic)
• Social connections to other users (structural
  equivalence)
•   
• Familiarity and Novelty 
•  
• Past collaborations
• Structural equivalence/distance

14
How Does Team Composition Influence Outcomes

• Are there correlations between these team
  measures and outcomes measures?
• Team size vs. ratings
• Tools 0.11 (p .43, n57)
• Team size vs. of users
• Tools 0.28 (p lt.01, n129)
• Publications -0.12 ( p .43, n45)
• Team size vs. of jobs
• Tools 0.017 (p .85, n129)
• Team diversity of universities vs. ratings 
• Tools 0.07 (p .62, n57)
• Team diversity of universities vs. of users 
• Tools 0.37 (p lt.01, n129)
• Publications -0.13 (p .40, n45) 
•  Team diversity of universities vs. of jobs
• Tools 0.20 (plt.05, n129)

•  

15
Team Composition vs. Outcomes

• Poisson regression
•    log(E(outcome)) log(duration) a
  team_size b  team_diversity  c

16
Future WorkConstructing Diversity of Research
Interests

• Given the specialty of each team members,
  calculate the degree of similar research
  interests in a team
•   Raw information on specialties 
• Tags pros--users provided cons--incomplete infor
  mation
• Subject categories pros--standardized in WoS and
  more papers included
• Keywords pros-fine grained measure, use paper
  abstracts
• Diversity measures 
• Unique count unique specialties/ total
  specialties
• Pairwise similarity Average(members i and j'
  shared specialties/members i and j' total
  specialties)
• Blau's index 1-Sum(Square(member
  i's specialties/ total specialties))
• Weighted Diversity use weighted of specialties
  in the previous three mesures, e.g. tf-idf weight
  (term frequencyinverse document frequency)

17
Research in Progress

• How does the composition of NanoHub teams
  co-evolve with collaborative relationships
  outside NanoHub?
• Are team assembly mechanisms that we identify for
  tools the same or different as those in
  publications?
• Is there a pattern in the sequence of teams
  co-authoring publications and tools?  Does one
  tend to precede the other?
• Do individual roles of members influence team
  outcomes?
• e.g. Do cosmopolitan individuals spread effective
  ideas and techniques across teams?
• Does the relative composition of veterans and
  novices influence team outcomes?

18
Questions?...