We talked briefly about some comparative statics.

There's much more that's actually known in the literature than we have time for.

And here the idea is, you know, if you increase the density, if there's

complementarities, if you sort of increase the density between

relationships, if I care about the total number of people learning a language, and

now I'm connected to more people, I have more of an incentive to, to learn a

language. So making those connections more dense

gives people more incentives to take, undertake these actions, and so you can

get results like that, bringing in changes in the network to understand how

that's going to lead to changes in behavior.

Multiple behaviors, so when is it possible that you're going to have

different technologies survive, and different groups taking different

actions. Well, homophily, cohesion, segregation

patterns in a network, things that we talked about earlier in the course, come

back now to be very important in understanding whether or not you can have

different actions surviving in different parts of the network and that has to do

with how introspective different groups are.

We looked at a couple of classes of, of continuous action games and in particular

these linear quadratic games gave a very nice simple solution where the intensity

of behavior ties back to the position of a node and to centrality measures, bond

assist centrality and so forth. So that's a very tractable model, a

simple model where you can get a closed form solution for how people are going to

act as a function of a few parameters. So obviously, it, it's, it makes lots of

assumptions parametrically, but it ends up giving you very tight and simple

predictions of what's going to happen, which then can be taken to data.

And so these kinds of models are being used increasingly in understanding pure

effects and modeling behavior diffusion and things when people care about what

else is going on. So, we can begin to, to marry these kinds

of models together with diffusion models, learning models and so forth and we'll

end up with a very rich set of predictions for behaviors.