[MUSIC] So we've just completed study number one. Study number one was looking at the spreading of lead users through neighborhoods. Now, what we're going to do in Study 2 is we're going to look not at neighborhoods directly. We're going to look at actual individuals influencing each other in a social networking site. This study was conducted by some colleagues at UCLA and also at the University of Maryland. So let me go through the problems or the challenges first that this office face in trying to do the study. First challenge they face is if you're afraid of someone on a social network, their could be a very strong tie or it could be a very weak tie. That friend could be your best friend that you grew up with and you known since five years of age, or it could just be somebody that you met at the function and you exchanged business cards and became connected as friends on Facebook. So the first challenge that they faced and looking at these social network in sciences, what a friend actually means for some people could mean a very, very strong thing, for others it could be very, very weak. So they didn't want to necessarily look at connections among friends per se. The second thing that was kind of challenging, is these databases are huge and the number of connections is very, very large, so they had to think of some simple statistical way of trying to get at that. And the goal of the study was to try and understand who's influential for whom. So Chris and I are connected on a social network in site. Am I influencing him or is he influencing me? That was the goal of the study. So for those of you who enjoy a little bit of history, I did some digging around and it turns out that the very first social network in site at least in the United States, was one called classmates.com. Since then, at least in the US, we've seen many come and go. There's been Friendster, there's been MySpace and of course now there's Facebook and who knows, Facebook seems here to stay with over a billion people currently part of that community. So the researchers wanted to understand who in the network is influential for whom. So what they decided to do since the measure of just pure friendship is not that diagnostic. A frame could be someone that I barely know or frame could be someone that I've known for 20 years. So the way they measured influence was quite clever. What they did was try to figure out if my activity in the social network was influenced by somebody else meaning after they did something, I also started to follow their activity as well. So to go back to the example of Chris and I, let's say being connected on Facebook, he's going to be influential for me after he start posting content, and photos and videos, I start going to his side and start looking at it. I'm not Influential for him, if I'm doing those activities but that's not affecting his activity at all. In simple terms that's what the office were doing here with the study. So what did they find? Well, they wanted to try and figure out who was going to be important and who was not going to be important and on average how much influence goes on a social networking site. Now, if you think back to some of the terms that Pete mentioned on his part of the course probably of the word that he mentioned the most knowing Peter haven't done the exact encounters the word heterogeneity. Heterogeneity is one of our great buzz words in the marketing Coursera course just means people are different and we have to understand the extent of those differences. So there was huge variation they found on the level of influence that was going on. Some people were highly influential, some people were not influential at all, others will highly susceptible to influence, others will not influence both people at all. So what are those numbers kind of look like? Well, here's the bottom line from the study. The office found that on average you are influenced by a about 20% or 1/5 of your friends on Facebook or LinkedIn down or whatever of the social network you're active in, about 1/5 of them are influencing you, and the other 80% or so are not really having much sway over your behavior at all. Now, if we turn the problem around, this statistic to me is also very, very interesting, they found that about 1/3 of the people in the social networking site were not influenced by anyone. These are kind of the maverick people who just do their own thing and they don't worry too much about who's posting what and other things that are going on in the social networking site. So you're influenced by 20% of your friends. About 30% of you out there are not influenced by anybody. Now, let's sort of dig under the hood a little bit and try to understand the extent of variation in influence and what causes influence in a social networking site. So let me now explain the blue histogram that you see in front of you. This is just a histogram taken from the original article. What it's showing is the amount of influence that friend f has on user u, and what you can see towards the left-hand side of the chart, is there are many people who's influence factor, if you like, is very, very small close to zero. And then, the right-hand side is a little bit like a long tail diagram again and the right-hand side, that the extremely or there are some people, a smaller number who are hugely influential. And on average about 20% of the social network people in the social networking space are influencing other people. So what you can see on the screen now is another chart from the paper, and I'm just going to explain the key results here. I think these results are actually very interesting, actually fascinating results, and things I think that we can not only use but maybe also relate a little bit to our intuition. So the first thing the office found was, someone who's been on a social networking site for a longer period of time on average is more influential than somebody who's just joined. I think that makes sense. That's a nice statistically significant effect. The second effect, which I think speaks to cultural background, as well as ethnicity, is that people who are from the same ethnic or cultural background on average, have more influence over each other than just random people. This is, again, partly due I think to homophyly. So I'm more likely to be influence by somebody who's from the Australia Museums kind of part of the world than somebody who's just coming at random. The next thing that I look at was gender influence. Now, this one I find particularly fascinating. So of course, there are two genders and four possibilities for influence. Man could influence woman, man could influence man, woman could influence woman, or woman could influence man. Out of those four possible combinations there was only one statistically significant path of influence the guys and the girls out there can probably relate to this. Girls were influential over guys, but not the converse. And again, think about what the definition of influence is in this case. The definition is when somebody is engaging in activity in the social networking site, posting, commenting and so on, other people are checking that out and following along. So when females do that, males follow along, but not the converse. So there's actually a lot of interesting research that's being done in the area of gender segmentation on the Internet and I think this is just another find in a place where into that. The final result that they found is to do not what who you are as a person or how long you been on the site but what it is that you talk about and what you say when you get there and how you present yourself. So we've already discussed reputation and review. This is a little bit of your personal reputation. It turns out if you're on a social networking site and you're indicating that you're looking to date other people, that significantly reduces your influence. So maybe think about that before you start posting too much. One of the implications of this if we want to advertise on social networks or we want to run social networks and so on, there’s really three things that I put on the slides. But let me just go through them. First of all, simple counts of whose a friend of whom are not really sufficient to understand influence because sometimes a friend can make a really, really close friend, other times it's a person you just met that's really too much variation. So we need different ways to measure who's influential on a social networking site. Secondly, the office found when they did some simulations if you take the very best people out of the social networking site that dramatically reduces the value. So just like in the real world there are some special people who have disproportionate influence over others, that's very, very important to keep in mind. And then, the final point that's related to the one that I just made, is that if you want to advertise on a social networking site or you want to use a social networking site to promote the products and services that you may be wanting to offer to people, most of the payoff you get is from identifying the very best and most influential people. Since many, many people aren't influential at all, there's great returns to figuring out those who are the best in this environment. [MUSIC]
