So, for the final model I am going to discuss tonight, I am going to talk about what many consider to be one of the first agent-based models [to], especially one of the first agent-based models to really explore an important social issue. And this model was originally proposed by Tom Schelling who would later win the Nobel Prize for his work on the Cold War. But he originally published this model in 1972 to discuss the issue of segregation. So, the model is alternately called Schelling's tipping model, for a reason we will discuss, but also the Schelling's segregation model since it explores that particular issue. So the way the model works: it is often a good place to start if we hit the setup button. You will notice the world is covered in red and green squares. These squares represent people, in particular people living in houses. Right? And the idea is that this is in some sense some fictional representation of an urban pattern of living. The red and the green squares represent different groups of people, you know. Traditionally within the American urban landscape this has been black and white individuals. Right? And the idea that Schelling wanted to explore was: Was segregation of these groups of individuals a product of people really being racist, [are either] they refused to live near people of the other group, or was it simply the result of a lot of individual actions? So he created a model with a very simple rule, which is that as long as 30% of the people around you are similar to you, then you are happy, so you do not want to move, you do not want to change your location. If, however, more than [30% are unhappy, sorry] 30% are different from you, then you are unhappy and you do want to move your location. Right? So, in this world which I see, anyone that does not have an X on them is happy, anyone that has an X on them, is unhappy with their current residential [preference] location. Right? Now, it is important to know, right?, that Schelling chose this role for a very important reason. Right? 30% is saying that I would actually be ok with living near people [who are] 70% of whom are different from me. This is what one, you know, might call mild preference or slight bias to being near somebody like you. It is not what many people would typically call a racist bias, that says that I refuse to live near someone who is different to me. Right? And so, what Schelling was interested in was: Will a bias like this, a bias that was small, not very high, still result in a pattern of Segregation? And so in this particular world we have a couple of inputs: we have a slider that controls the density, we have have % -similar-wanted. And what we are going to look at is, when the %-similar-wanted is 30%, at the end of out run with everyone has moved. Right? Is the %-similar [?] similar to what it was at the very beginning? And, of course, because we randomly assign them, it is roughly 50% at the very beginning. When we [?] assign the housing locations. Now, one small fact I have not mentioned is, how do the agents decide where to live. The answer is quite simple: they just move to a random open spot in this model. Right? They are not actually trying to find the best spot for them to live, they are moving to a random open spot. So, enough talked. Let us set up and run the model and see what happens. So we already set it up. If we hit 'go'. What you see is that the model settles down eventually, now everyone is happy. But it settles down to a point where clearly we seem to have a more segregated pattern of living than we did when the model started. In fact, if you look at the dynamics of what the %-similar was (the average % similar for all agents in the model), it grew over time and actually peaked at around 76% of the number of agents. Right? And so, given the simple settings, right? In the simple, you know, low bias towards people who are similar to yourselves, you still get segregated patterns of behavior. And so, Schelling, kind of, clarified that you did not need a large racist bias within a population in order to get segregation, but rather that segregation was what he would later call a macrobehavior that was resulting from a set of micromotives. Right? In fact, he wrote a book of that title 'Micromotives and Macrobehavior' in 1978 that recapped many of these findings, there in this paper. Also a small side note, an interesting side note: Schelling in 1972 did not have access to the computers we have today. My little laptop is able to run this model many, many times in a few seconds, essentially when I am after as much as is possible. Schelling did not have that luxury, so he actually did all of his work with a chequerboard and pennies and dimes, manually moving the [chequer, sorry] the pennies and dimes across the board to calculate the statistics that resulted in his paper. So, it is very intriguing, you know. Agent-based modeling [is], imagine the past, usually involves a computer, a computer simulation, but it does not have to be, right? In this case we had an agent-based model, for Schelling did not call it that term, that essentially was an agent-based model, that was done with nothing but pennies and dimes. Very interesting to think about. So, one last thing: Why is it sometimes called the tipping model? Well, if you let the model run, and we let it run several times, you see it almost always produces the same patterns of behavior. Right? And the reason why you call it a tipping model, the reason why it is often called a tipping model, is that if you look and you slow down the model runs, what you will see [when you let it run] when you run the model, is that when many of these neighborhoods are actually quite happy, there are groups of individuals living happily together, what happens if one particular individual moves in or another particular individual moves out, and that causes the neighborhood area to become much less happy. And that tips the neighborhood from an integrated neighborhood into a segregated neighborhood. As a result of that this is often referred to as the tipping model or as the tipping point that occurs within that system as those individuals move around. You can also play around, by the way. There are several other parameters. You can see if you increase the density [what is] how does that effect the model result. So, let us try to slide that you can see it faster. And, what you will typically notice is that as you decrease the density, the amount of eventual segregation that occurs, goes down a little bit. But it does not go down as fast as you would think. And that is partially because the %-similar factor is measured based upon the people who are near you. Right? You can also manipulate the other sliders. So you can also play around with how many %-similar near you is required to be happy. And if you do that you find that if you decrease that slider even for a little bit, you [get] actually you can maintain this integrated neighborhood. Right? And so, that 30% value, in many ways, was a key value. Much less than that, and you do not get much segregation at all, much higher than that, you actually get a different problem, you get a world in which no one ever [are] quite is happy in some cases. And this case actually it looks like it you might settle down at this particular place[?]. But, maybe not, actually. Looks like there is a border. But you get a very, very segregated pattern of behavior in this particular case. Something sometimes I refer to in a lecture all to[?] segregation. So I encourage you to take a look at these models, play around with them. One of the important things about agent-based modeling it that often [you don't] you do not necessarily develop a [model] new model completely from scratch, instead you are inspired by another model that you see. And you want to copy some of the dynamics, some of the interesting ways adopting the model works and then use them in a new setting. So, having a literacy about models is very important in agent-based modeling in order to create the next great model you are going to work on within your particular area of interest. Right? And so I highly recommend you, play around not just with the three models we have talked about today, but also with the rest of the models in the Model Library and any other models you can potentially take a look at.