So before we summarize this lecture, let's point out 5 considerations. Based upon the analysis we just did. The first thing is that, we've assumed that everyone acts rationally. They can decide what the best guess to make is depending upon the current information that they have. But in reality, what each person should do, can be quite different from what she actually does. Researchers have observed for instance that, if you play out the number guessing experiment, it's not going to play out in reality as the theory predicts. And, you know, the reason for that is that we're probably not all able to go through all the probabilistic thought processes on the spot to figure that out. And especially if we've never heard of the game before. Also, how many people are needed for a cascade in general? Well sometimes it takes the apparent conformity of more users. You know, like 10 people instead of just 2 in, in the case of the number guessing experiment. Especially if it's something with a low probability. You know, like for you to believe that there's something wrong with the sky. There's something wrong enough that you would actually want to stop looking or stop walking and take a look. In this case, you'd need more public actions to override your own private signal. So, the stronger instinct you have, the more people you're going to need to see to know if you'd over ride your own instinct. So this going to depend upon two things, right? One consideration is going to be the task itself or the event itself. [BLANK_AUDIO] So, you know whether it's a number guessing experiment or if it's believing something's wrong with the sky, and the other one's going to be the person. Because some people just have stronger instinct than others, especially if the person's very busy. Not going to want to stop and take a look. Another consideration are the implications of this information cascade model. So, there are many implications because various social phenomena exhibit the features of a crowd making a decision that ignores each individual own private information. At the same time, that chain reaction is going to be fragile. So that's why, you know, in certain circumstances like the, the fashion fads, stock market bubbles. This can help to explain why so many people, will, will congregate and do the same thing or follow the same public action. And why once people suspect the underlying true signal has changed, whether it actually has or not, that the cascade can quickly reverse or just dissolve. The next consideration is, how does this translate to viralization? So how do we take the sequential decision making and information cascade and translate that to viralizing YouTube video? Well, it's not easy. There's definitely a theory of practice gap here. But the main idea should be clear. You want your video to undergo an information cascade, so that when a person sees or hears about it, which is their public action, they will most likely watch it, irrespective of whether or not the person's actually interested in it, which would be their own private signal. So, if you have you know, your, your YouTube video. Basically your going to want to, instead advise people to come watch your video on your page. Whether it's through recommendation, or whether it's through subscriptions or so forth. You're going to want people to come there and you're going to want them to come there whether or not they have an in, and intrinsic interest in the video or not. But how many public actions does it take before a [INAUDIBLE] person will watch a video automatically? For instance, how many total view counts does it need to have before someone's going to watch it without looking at their private signal? Does such a number even exist? And clearly it's going to be different for different people. Depending upon how mailable the person is, how willing the person is to just follow the crowd in the first place and so forth. All of these are open research questions with no definitive answers. The last consideration is that so far in this chapter we've focused on a single influence model, being information cascade. And we've used this to show how people's actions may be dependent upon others. But what we have assumed is that everyone has access to all public actions before them, as shown in this diagram here. So, if there's 4 people and A goes first, A doesn't have any public action before him but B will have, be able to see all of the action from A. C will be able to see the public action from from B and from A, and then D will be able to see the public actions from A, B, and C. So this is a population based model. Meaning that it's assuming that population interacts as a whole, rather than taking social relationships among people into account. So there's no account for a specific social relationships, which are naturally very important in studying the diffusion of information, whether study of influence because you know, depending upon how you feel about the other people. If the person's your best friend you may be more likely to follow their lead than if you don't know the person at all. So in the next lecture, we're actually going to study relationships that are topology dependent. So topology dependent, which means that they take the social graphs into account. [BLANK_AUDIO]