Figure 1 helps illustrate how netwo

Figure 1 helps illustrate how network diversity varies as a
function of the two components described above. The two
clusters in the figure are knowledge pools. For the purpose
of illustrationt,h e top cluster is chemistry and the bottom
cluster is mechanical engineering. Solid lines indicate strong
ties, and dashed lines represent weak ties. Connections are
stronger among the chemists than among mechanical engineers.
Chemistry and mechanical engineering represent two
distinct areas of expertise, so there is a primary hole
between them. Disconnects inside an area, secondary holes,
indicate internal heterogeneity. The stronger ties among
chemists indicate less internal heterogeneity than in mechanical
engineering. Strong network connections, or the absence
of secondary holes, inside an area indicate the absence of
diverse knowledge (e.g., chemists have less diverse knowledge
than engineers). Therefore, increasing Pk indicates the
absence of diverse knowledge inside a knowledge pool The second element of network diversity is how an individual's
network connections are spread across areas. The two
individuals in figure 1 are exposed to the same contacts, but
Person 2 has the more diverse network. Person 1 has a
stronger connection with chemistry than mechanical engineering.
Person 2's network connections extend equally
across both areas. Network diversity increases as an individual's
network connections are distributed more evenly across
multiple areas of expertise and ties among people in the
same area of expertise are weak.
Although network diversity and network density are distinct,
they are not mutually exclusive. Density around a relationship
exists when two people are strongly connected by mutual
third parties. But the network can still be diverse if it includes
people from different areas of expertise (e.g., engineering,
biology, mathematics, etc.). Accordingly, network density
does not come at the expense of network diversity. In the
current analysis, network density measures the presence of
strong third-partyti es arounda connection, and range measures
the distributiono f connections across differenta reas of
expertise.
Structurale quivalence. Finallya, s another indicatoro f common
knowledge, we used the network data to define the
extent to which two individual occupy the same network
position. Individuals ho are involved in the same pattern
connections are equivalent and can be expected to have similar
knowledge and information We measured the degree of
nonequivalence between the network pattern surrounding
the respondent and the pattern surrounding the focal contact
as The measure is the Euclidean distance between two network
patterns. The distance is high when two individuals interact
with different colleagues in the organization.