In this matrix, you can see the friendship relations during the semester in which I observed them. I used various network analytic software packages, to identify the groupings, and how they broke down. In particular, there is four cliques, or four clusters. And, you can read the ties within and between these clusters, By following the row column relationships. Or along the row as a from to relationship. Hence the value of one in student number 16 to number 15 is a 1. And there is a dot from 15 to 16, and that suggests that 15 thought that she if friend's with 16 but 16 does not reciprocate that sentiment. So, notably most of the groups are hemofilus by grade, gender and race, so they follow the saying of birds of a feather flock together, moreover, many of them have reciprocated ties at least more so than by chance. While it's not shown it's also the case that many of these friends sit by one another, so perpinquity is also an effect. In the matrix, you can see the smaller secondary peer groups underneath the larger core grade level groups. And you could see that they're hanger ons by looking at the off-diagonal relationships between those two clusters. These secondary cliques seem to want to be friends with the larger core clique within their grade level. But it's not always reciprocated. And because of this, there's kind of a rank ordering to their clustering of cliques within the class. We can render these relationships into a network image where the y-axis is the prominence or popularity of the individuals, and then she shaded circles reflect general boundaries of each clique. Notably, we see that the two grades are somewhat disconnected and each having a core clique with a hanger-on, just like I described earlier. In other analysis, I tested whether observed interaction patterns conformed to these cliques, over and above the seeding and homofily effects, and they strongly do. For easy of interpretation, I'm just going to superimpose the observed behaviours interactions on these groups. From doing so we learn a few things. First, we learn that most of the interaction is directed within the cliques. Second, we notice that the cliques specialize their behavior. So here I'm using red to denote where task or academically focused interactions emanate. I render the red color bolder where the rates and densities of that kind of interaction are higher. And here it clearly shows that the core senior clique dominates such interaction. And the core clique in either grade is slightly stronger, so its status is important as well as the kind of group. Next, I use blue to denote where social or non academic interactions emanate. Like play, joking around, things like that. I render the blue bolder where the rates and density of such interaction is highest. Hence, the junior core, group dominates that kind of interaction. And the core clique within each grade actually dominates that interaction. So you can see the transition here. The senior group dominates tasks. And then the junior group focuses on social affairs. A slew of statistics can accompany these images and further the argument. However, the point for this lecture is more conceptual and schematic. The structure of the informal network and it's cliques. Strongly guide behaviours, that's the point, moreover, the cliques arise from tie formation mechanisms of same homofily, reciprocity, status-seeking and even an effort at specialization, so that they avoid competition across the groups. So, that some of this is that it's not just single relationships that influence workers and their firms, but also network positioning and the groups within them that shape kind of outcomes and worker behaviour. But what of network formation? In the introduction, I spoke of how analysts often view networks as an outcome, or as having a desired structure that managers want to achieve. How do we accomplish and engineer different structures of association? In my own work on highschool adolescent networks I found that they vary in macro structure from school to school and from classroom to classroom. Some of these settings entail hierarchical worlds like the first image to the side and others that are heavily segregated and clustered like the second image. Nevertheless, friendship networks are all shaped by the same sorts of time mechanisms, and those mechanisms are related earlier as homofily, birds of the same feather, flocks together, Or propinquity, friends of convenience. Reciprocity, I give you something, you give me something back, and high creolization. A sense of dominance and control. This kind of raises a conundrum. If the same micro-mechanisms apply to every friendship network, then how is it that their patterns vary? How is it that we see these varied macro-structural environments? The potential answers are interesting. It turns out that certain types of ties usually correspond with a type of network form or pattern. In the case of our schools the ties are friendships, and friendships tend to be reciprocated and local so they accent clustering more than they do rank ordering. And this closure and reciprocity of friendship ties happens to be the strongest feature driving high school friendship network formation. By contrast, if the friend's ties actually weren't friendship, but weak ties, then it's likely the structure will entail more spanning trees, rank ordering, and fewer groups. Acquaintance ties have greater imbalance and looseness to them, and they enable different network patterns to arise. So depending on the type of tie, we might expect different kinds of network forms. But this doesn't explain how friendship networks differ across contexts. In my work at high schools we find that the friendship networks vary from school to school because the organizational context amplifies and dampens the salience of certain micro-mechanisms. What this means is that the compositions of participants and the utilized organizational rules moderate natural basis of association. Take the case of a large heterogeneous population of say, multiple equally present races in a school. And one where contact is by choice. And this means that the students aren't sorted into ability groups by the school. In these kind of cases, we find the pattern of association is the most highly segregated by homophony. They have ample opportunity and ample variation and traits to do this. Now, in large schools with no organizational sorting and lots of choice there also tends to arise rank ordered clicks. So not only in those large organizations do they form clusters of certainty and trust but also some degree of ranking. The only time we see random dense associations are when organizations have homogeneous populations and they place members in small interactive settings and they sort them beforehand so that the differential characteristics of these students are already kind of sorted out. And in these contexts we see if we just induce kind of greater interaction, that they have kind of a random mixing. The same occurs with small settings like classrooms where you introduce rotating group memberships with frequent interaction of these individuals what happens is that you can induce a kind of of collaboration structure that's highly dense, but doesn't have these kind of cliqueish hierarchical structures. In addition, if you keep people focused on discussions about tasks or the materials at hand like organizational analysis, the form relationships on the basis of those contents. And so, already we have some sense from these kind of these models of network formation of what kinds of treatments would induce the networks we d like to see in organizations. Once analysts have a good description of a network and once they have some sense of the key influence processes on worker outcomes and, once they know the key mechanisms that drive tie formation to assume certain patterns, the analysts can begin to prescribe all sorts of treatments. Now, it's going to differ by context, and by question for many of you. But I hope, a lot of you inferred what these treatments might be, from the prior slides of this lecture. Nonetheless, what I can do is tell you what many firms will probably ask you for. They're going to want to facilitate the creation of efficient network patterns. And for example, many of them will want interactive, dense networks of positive work related collaborations as opposed to positive sociable collaborations that don't relate to work. And they're going to want those networks to spam groups, so good ideas can travel around in the firm and across them. And many companies will also want to forge teams composed of differently skilled persons who are going to rely on each others' strengths. So, they're going to want an organic whole that's greater than the sum of their parts. And this Will supposedly make a product or solve a complex problem. So they're going to want and seek these kind of things. In addition to facilitating the creation of ideal network forms, companies will also want the analyst to perform network corrections or to solve coordination problems. That are endemic to the networks already in place. Now, within schools, we have many instances of this occurring where students assume positions or they form groups that drive behaviours in negative directions an underperforming group. For example, in many classrooms certain kids dominate and take up all the teacher's time and attention. Thereby making achievement gains less equal. To offset this, researchers have suggested positional treatments. To offset, inequal access, the tasks are designed to involve decentralized formats, like group work. So more people can talk, and then they called for differentiated roles so everyone has a job to do, no one's left out. And this type of treatment equalizes status within the group and it renders participation more active and even so that knowledge acquisition and attainment is similar. Another problem in schools though concerns group norms and peer influence through cliques. To offset this, scholars suggest propinquity changes. Such as rotating groups and seating assignments. Things that we've all seen before but which we haven't managed manually enough to overcome those kind of sub optimal solutions. So these are kind of a simple solutions in a certain way, but I hope they help you think about how networks not only can be formed, but also how they can be redirected through various managerial efforts.