 to all coming here to hear me talk about my Background is in mathematics and I spent last 15 years working in computer scientists I did teach mathematics computer science for five years. So actually I know it means having students in the classroom They are all the thrills and the pain Sunday. It's all together. So in Microsoft research in fact Yes, it's true We have 900 researchers across the globe and very very strong PhD program So we do get lots of students that work with us But there are other people students who come to work with us as interns So we do have a very broad exposure to the younger generation and their thinking so presentation that I sort of composed today and It's sort of inspired by some of the talks that I heard Earlier during the conference and I was trying to find an ankle that would particularly be useful For us thinking about education and learning The work we normally do is computer science work It makes the research is a laboratory looking forward So very much away from any products what we call is Undirected research. So what you see here is effect a set of a computer science disciplines that would be in any other University One particular area here is called computer immediate living and that's the group I work in and my team is called Integrated systems is completely cross disciplinary. So we do have people who are working on user experience design architecture the whole spectrum and I thought perhaps to share with you a couple of things to get an insight how digital technology has transformed our lives and Penetrated all Experiences in I picked actually four examples because one is particularly important understanding how children actually Want to bring technology into their magic world. The first one is called tell table where they want to use the latest technology Then they play with in sort of education settings. So I'll tell you a couple of words about the tell table then Something that's particularly appealing to students who would like to get hands-on and create something really new Really real. So we all use digital cameras But people don't really know how things work and how would the program a digital camera So the gadget theory example. I'll give you is exactly about How do you put things together? You can program them and then feel very good about it because you build something that actually works and Anybody ask you about the latest technology coming. It will be about Body gestures. It's not about multi-touch anymore. It's not this anymore. It's that way be hands in front of the either game console or yes, you will see medical equipment and Finally, I just want to tell you briefly about Research in actually collaborative learning that we've done because I heard lots of talks about how do you bring technology in the learning place? And it seems to me that understanding the ecology of it might be helpful So a couple words about the tell table I'm not sure whether how many of you have actually seen some of the YouTube videos But it is about enabling children to bring physical things into their magical world in this case is the Multi-touch table they can take pictures of their favorite toys and they can distort them They can create any sort of things that you can do in the digital world Cut up pictures put the face on the top your teddy bear teddy bear that looks more like a lion So these are the examples where actually they created Assets and then they are moving them in a digital space and telling the story and most importantly they can record in themselves And put it on YouTube That was by the way the latest request for my youngest son. He wanted to Videotape himself about playing the PC game and then put it on YouTube said everybody can see This is on a baseline these days on the on the dot-net gadget here. This is basically a set of Harder components that can be ended software development toolkits that enables you to create and Very quickly prototypes of devices and sorry and one particular one that is being shown is Abort with elements that you can create a camera Connect is the latest it is a Adapt camera that enables you to do the gesture recognition Sorry, this is kind of moving without my control and now it's being controlled Now it's being considered in medical field because in an environment where doctors cannot for example Use their gloves to touch things. They have to instruct the computing environment to the gesture As you know, this is the latest greatest part of the gaming industry that people don't really know use any Retonsals anymore. They want to be able to instruct computer with their gestures and this example and I would like to talk about is about a research about learning and research practices at the Computer laboratory in Cambridge that we have done in collaboration with the Jeremy Bamberg a professor on nanotechnology A lot of discussion here have been how do you how do you bring technology into the learning and And what do you do? When you're transferring from so paper paradigm into the digital paradigm What happened in nanotechnology? It was very easy to get of get rid of paper because all the experiences with a Physics and material science of nanotechnology is so abstract. You can't see anything anyway You hear see the equipment a very very complex and Sophisticating equipment that enables scientists in this case actual students and you also came to see how the materials look like By shining the laser lights and based on a software Providing them with an image of a graph or what they have there They're kind of thinking what they can see through the digital technologies So there's no direct human experience here. It's all mediated through technology and Therefore for this sort of research is extremely important to move things very quickly From the data to information to the conclusion to the findings and in the end published paper The first thing that a professor Jeremy Bumberg wanted to get rid of was the paper form which would then Because it's not they did native digital would interrupt this very fast law information So he was considering how to replace what they call a lab book And as you know in every research environment lab book is extremely important is the way we capture observations and the way We store our knowledge and transfer it now In order to facilitate that he introduced as I replaced in the digital form in this case In this case he just repurposed a software that existed on the market. I think this one is actually one note So you can record everything you've done now What happens is it's not as simple as that replacing one single artifact is actually created the whole workflow So when Jeremy needed to make decision of what sort of a digital Software and how that he would bring in the first thing was about the practices in the workflow It wasn't about any particular device It was about what needs to be accomplished and how that could be done in both efficient way So I'm just here showing you a workflow that happens in that in this environment In which he needs to teach students how to run experiments then need to verify that actually experiments are properly done and then in the end they need to decide what you're gonna do next and It results in extremely complex ecosystem of devices and content So here I'm showing the content that comes first when students are recording experiments. That's for say one note And then when I need to discuss things This is a single PowerPoint slide and professor ask what is the point of this experiment? can you point me to the data and finally when they actually have a joint discussion and Make a conclusion and plan for the rest. I'm bringing this up because Decisions are often made on an individual basis, but in fact it is about understanding the whole Digital environment that it's going to incur and how this will be supported because digital is the most vulnerable of all the media It is the most vulnerable if you turn off electricity. You can't see it Now even if you can if you if your electricity is on and you can see the files if you don't have appropriate application to look at it You can't see it because we humans cannot directly consume digital So understanding the environment is absolutely critical for our knowledge for education for our future And there's the whole area about how to preserve the digital and How to ensure that future generation can benefit from what we know now And I brought it in this context because in this particular learning environment It was understood that the benefits of digital which we should take advantage of are fantastic a very fast Aggregated data distributed knowledge shared perfect But the next but the thing is about cleverly putting it together to support the workflow Okay, I hope didn't distract you too much from the main message We are going to move now into network analysis. It really ties very very tightly with this a Concept of it of a digital because the way the reason why we are nowadays talking about a network analysis is because the digital technology has enabled humans to Collaborate and communicate at the scales that we have never experienced before It is really the digital that enabled that so What I was going to talk go over is slightly a brief history of how things evolved with the computing technology how we will move from first from the computer networks to the engagement and then I Talk a little bit about the tools that are now needed for us to understand what is going in these new social ecosystems and Hopefully get us to how we can teach the next generation or perhaps it learn from the next generation How we can think about these networks and prepare for the future okay So I picked this slide because it is in fact from 2004 We started looking at the networks and somebody said, okay, why don't we look at the? Discussion groups online because that was a realm of computer scientists Computer scientists would mostly put come together and discuss the technology Form the news groups and this is indeed how it looked at it and it looks now as well So these are the computers that are connected with each other with special protocols so you can send messages And what happens and so on top of it you have a you have a client Interface through which individuals can exchange messages as I said before digital doesn't exist unless We have a client. We can see things Now this is then enabling interaction What normally happens there is infrastructure and there is a content there's an engagement in the content So in this case at that time These are at the time really staggering numbers There was hundred ninety thousand different news groups of different small communities And they were exchanging many many messages seven hundred thousand messages a day And at that time they were identified more than nine million users and at that time as I said this was a really really Fantastic and staggering numbers on top that there is an engagement Network that we can start looking at so who is communicating with whom now that was the first time That's sociologists actually had data about the human interaction So it was a fantastic opportunity for them to start looking at the whole environment and truly very quickly what happened is starting analyzing the such ecosystems And we arrived at the conclusion that it's almost like any other system There are people that have to satisfy different roles for this sort of Ecosystem to work. There is somebody who is going to support the machines So this is the person who is investing in the machines to support interaction Then there are people who are going to be managing it. There are people who are Monitoring it researchers and then people who are participating So it is very very complex ecosystem of different roles Now among participants in particular and let's let's imagine ourselves in In this sort of net communication network, there are people who will be doing different things and You know, you may be belonging to different Online communities and may assume different role in different community when I first look at this I was trying I felt very guilty because I am the one where they call Silent searcher or non contributor most of the time So there are people who are leaders in the community. They're really really active There are people who are asking questions and leaders that the ones will be answering or Answers would be once we're answering and there are people who are just watching what's happening And if you ever feel guilty about please don't because in almost any community about 2% of people Contribute everybody else consumes okay, so These were the kind of general roles that Sociologists is where we're aware of and it was very nice that now we had data to start looking at it more detail Now before we go into the analysis practices, let me just move you closer to the Reality so that was in 2004 2012. What do we have now? There are new services appearing all over the place and all these services are in fact created to satisfy some Human needs now people like to socialize a Facebook is the perfect place it's a place where people go and They can communicate with their friends that can accomplish a lot of different social needs then What what are the numbers we're looking at now half billion people involved? Now if that scale is not enough, let's just look areas that are perhaps not necessarily now main A focal point. I know about this mostly because of the children my own children So here's this another another service. It's a steam website where people go and play and And if you just look at the level engagement now between 1.4 million and 2.6 million at any point in time Are going to be using this Infrastructure and and they're engaging and it's not just about playing on the right hand side. You can see a chat In my almost goes with it There is not only chat, but there is also grouping there are there are clients people play they belong to a group of people who have an accomplishment Based on their Playing experiences that they get different badges they they get different status. So it is the whole whole world It's an imaginary world, but it's very very real for them And now if that's not enough, this is sort of still another picture, but a younger generation This is my youngest son. He is not typing. He's not chatting. He's talking He's talking and that's basically the minimum now criteria. You just put your earphones and you have a Game going on on one side on the other side. It's a live chat and that chat is not anymore to typing is to talking so you have to understand what The infrastructure that's supporting this sort of engagement and it has to be an excellent You know audio it has to be an excellent response of the system when they're playing and These children are now multitasking and doing from my perspective impossible things So Looking at these technologies and when they brought us the question is what are the social networks and How would we start thinking about them if we talk to sociologists? They will talk about ties among individuals if we talk to computer scientists They'll start thinking by the graph structure and the challenge for us is to bring these two together And this is exactly where the objective of my talk is to just explain how we are now enabling sociologists who are domain experts and a computer scientist who have technical skills and Very various in digital after all the digital came from the computer science So how these two needed to work together in order to enable for sociologists And then perhaps everybody else to start thinking about the social networks Okay, just To motivate a little bit The particular tools. I just want to mention one thing You saw specific examples of social networks and you understand now perhaps that It there is a extremely sophisticated computing infrastructure involved now this computing infrastructure Has to be somehow sustained. So somebody has to pay for it. Somebody has to maintain it If there is no engagement on the top of that infrastructure the whole system collapses Okay, so building social network services is an extremely difficult problem because there's a hardware and there is a human engagement and if either of them is Not performing as it should the whole system collapses If the hardware is not good enough to support X million people Interacting at one given time then the system collapses if there is infrastructure But very few people engaging they cannot sustain the infrastructure. It can't pay for it Okay, so it is extremely important for people who are building these to understand what's the healthy Relationship what is the healthy engagement among the individuals? So that brings us to the question. How would we characterize the health of the community and This is the same question that you will have if you for example decide that you're going to use some portal or some Engagement even chat room or forum for the students Understanding how to first create user interface to get the engagement going After that's in place to start understanding Why certain topics are being discussed and others are not why are students some students engage some are not so This is what we are talking about. We are talking about metrics to characterize the engagement and that's basically is the bridge between Techniques of analyzing the data and then interpretation Now lots of people say I Get it often but Natasha you stay away from any formulas or any graphs. Okay, so this is I learned this Okay, no speeches with formulas and graphs now My argument here is the graphs are becoming Quite popular and I'm just trying here a net for a graph that a colleague sent me this is his social network representation of Facebook friends and This graph is created by taking pictures of the friends that appear on the same photos in Facebook So I'm just thinking wow. This is fantastic. I mean if now everybody gets used to looking at these graphs Then we can have a conversation about them. Yeah good so About almost now five years ago We started a project and called node Excel and the main objective was to think how can we bring? graphing to the same level as people used to charting and Idea came okay Well, there is a one tool that we can extend and he's doing charting already And there is it basically a spreadsheet in this case is Excel So if you have a spreadsheet and you know how to use a spreadsheet and you know how to create a pie chart or Histogram in your spreadsheet you should be able to create a graph. That was the reasoning so Then we decided to extend Excel with a new little ribbon that would enable people to create these charts Okay from the data so to make it easy you can import Different social network data So if you if you own Facebook you can import your friends if you are on Twitter and you're following a hashtag You can download information but the hashtag that all goes into a spreadsheet and then you can click some buttons And then the graph should appear so this is the idea and this is actually what we've done the idea was Not everybody needs to know programming in order to visual graphs Not everybody needs to know programming to be able to identify their role as it's seen in the graph Okay, so indeed if you import some data, this is a Twitter data that from one conference that Microsoft held recently Which just shows connection of people who are tweeting I'm bringing this up just to show you what sort of things it will do So that you can import the data Then important thing was how to teach people Metrics and I'm just bringing here some of the concept graph metrics They look now like, you know, perhaps like Chinese it is closest centrality between the centrality eigenvalue centrality. These are all the Properties of the nodes in the graph that one can Discuss alone in the basic thing is Network is nothing else but set of these nodes that the entities and edges and everything else is just counting Counting the edges in different ways So it's not a big deal all these have some formulas behind But the meaning is interesting and the meaning often is best Understood through visual rather than mathematical formulas So so what I was going to do is show you a couple of examples of how you read these charts and then show you how after we do applauded with students how students started learning it and Using similar vocabulary to express what I meant. Okay One interesting thing as I said before is about the roles of individuals So what is particularly interesting if you pick any of these nodes you can then look what is the Neighborhood of a node. So if the person if a node is a person and then you can see how that person is linked to other Individuals so you can for every person extract a small graph Which we call the egocentric network and you can start interpreting those Okay So going back to the roles that remember I mentioned people participate in different ways And so if somebody is a person who is answering lots of questions question is how would that be reflected in the graph? So this is one example This is a person in the middle and there are lots of errors pointing out. So if anybody's asking a question This person but basically would be the one who will be answering to people So if the edge means person answer the question to a person then this Person in a century is the one who is very Gregory very a person very altruistic and very much engaged Now if you see something like this in the middle It's very likely that a person is answering to other people But individuals that are connected this person also quite chatty and they talk to each other So so then you can see a very interesting sort of a slightly messy And then pattern where we have a person for maybe it's setting a topic, but then everybody else kind of start participating and Then you have a situation when things are very dramatic as you can tell on the right hand side So this is a situation where multiple topics and multiple people very strong leaders are involved and you have a very strong discussion group Okay, so this is just on the high level I want to explain how just based on nodes and and these edges We can start a reasoning and I'm using the word the reasoning. I'm not saying that we can It is very important to understand that this information and finding patterns is the way for us to kind of create Hypothesis we still need to test this hypothesis okay, so I'm not sure how many of you using flicker, but I thought if I show you an example of a social network Where you might be familiar with and to see how we can extract networks and this is very interesting Service because it involves on one side people who are Publishing the photos Then there are photos as items who other people can comment on And another thing is people can tag the photos. So I have three things involved I have humans who are interrupting and I have photos and then I have tags And I can create all sorts of networks I can create a network, you know, who is Talking to whom about the picture or generally about pictures or which tags are related to which tag if there are multiple tags per picture so so The net was not arise only From communication people to people it can be any relationship among any items So let me show you in this particular service because we have people communication tags and we have individuals We can create different Networks and I'm showing you one in which the link means that one person is Commenting on the pictures of the other Okay, so so if you're interested if we spend attention to your pictures and flicker Then you can draw a graph like this You import the flicker data and in Excel you can create it So the person whose account this is is my colleague Mark Smith Mark Smith is one of the leader of the not Excel projects. So Mark has lots of friends and He's looking at this and saying hmm This is their sort of group in different clusters and the one but person is particularly important Because this person here in the middle Heather Heather actually is a gatekeeper for the whole sub sub community of his network So in a way, but just looking at the placement of the notes and connectivity You can see which people are quite important as connectors between different otherwise Perhaps not that strongly connected Subgroups of your friends Okay, so I'm mentioning this because through visualization. I can now tell tell you that this person here has a very very high Between a centrality if you remember one of the keywords there was between a centrality Why because this person sits between two clusters and therefore is very very important person in Mark's network I mentioned before that the networks are there to set hypothesis and This is an example to illustrate this Again, this is a flicker network and some pictures you can see are connected because this in this case They share a tag so sharing a tag two pictures are connected if they share a tag So you create the network and I start looking at them and really anything you say may look like reading a horoscope Right? Can be any any pattern you can try to get out The thing is that it is important to understand where the data comes from Once you know where the data comes from then you can start correlating it with something that is outside the network But explains the patterns And I think this is important to understand pattern analysis within the network data is to investigate and Make hypothesis they still need to be verified the truth is not within the network the interpretation isn't in the experts head And it's it requires human intelligence Okay, so in this case it turned out that mark was in Las Vegas and Taking pictures and the way he was tagging them in fact corresponds very closely to the Strip in Las Vegas as you go from one hotel to another Okay Very good. So and the final thing I want to show you about the network analysis. I want to show you a very common technique and That's the the filtering technique. So you may plot all the edges in the network But then you may decide that you just want to look at the most prominent one So you're just looking at the most prominent patterns and I like this example is because it's very simple self-contained This is the voting statistics for the ES senators. So whenever two senators voted The same either yes or no on an issue. There's a link between them. Okay, so now looking at this looks pretty Busy and it's really hard to understand any patterns, but then I say, okay How about if I just look at the links or those nodes that have at least 65% agreement so at least 65% of time These two individuals said yes together or no together And suddenly There is a separation now. I can see a pattern. I see two big clusters Well, then I can add a bit more information to this I can add information of a party affiliation And I see very clearly now The red are conservatives and these are Democrats So you can see very very clearly now that these two Actually represent something that I expected from reality But what I didn't quite expect are the three satellite nodes right in between in the middle now remember this The metrics that I talked about these three individuals these three senators in fact Have very high between a centrality of course from the from the party point of view may not be desirable in this case Because these are the the conservatives that the more leaning towards democratic Agenda, okay, so this is a very simple example the You can very easily do this in in the Nordic cell You can just say well I'm not gonna I want to hide those edges that are not very strong and then the patterns up here Great all right, so Just a little bit about that about teaching these sort of skills the students we decided to we collaborate with University of Maryland Professor Ben Schneiderman and the Mark Smith myself The Hansel and we have been being involved in in which students at University of Maryland in deployment of this software to see how students will learn and Just just kind of to set the expectations once you once you import your network Originally, if you if you include the pictures and everything else and without doing anything else to the network You may end up with a picture like this. So first thing that we want to teach students This is not really a good visualization to try to analyze So we needed to teach them what they should be aiming for Okay, they should be aiming for clarity in the visualization and the Ben Schneiderman couple of colleagues there They in fact Provided the guidelines as a set of guidelines for students What should be the way to design a good network? So first of all, you don't want Vertices to overlap second if you can avoid Edges to cross that's easier for us human to visualize and comprehend. So number of these rules So we deploy the software and We're looking for different, you know, these are like homework problems students were taught Generally social network analysis for five weeks and for three weeks They were using node Excel. We gave them instructions three hours once a week for three hours, so they had practically about nine hours of using the software and they gave us the Reports they had to pick the data and get the data from somewhere else for some community They wanted to study they put it in the Excel and then they were analyzing this now What was striking to me is that when I was Looking at these reports at the end. First of all, they all really wanted to get very clean nice Visualizations, they could move any node by hand if they wanted and so on that was up to them But most fascinating was the language that they were using so in this case the student was collecting data about maintenance of a network at the list server community Okay, and and this is the communication among the individuals and he identified who was the admin at the time The question was who else could become an admin in the community of maintaining the list? That was the question. So the student was saying my initial hypothesis is a leader should possess a high between a centrality and high eigenvalue centrality So the student was in fact arguing about what the characteristics of an individual would be in the terminology that She learned only within the past three weeks But we are not talking about an expert This person has never seen the formula for between a centrality or eigenvalue centrality, but can visually Identify these after being giving examples and not only it wasn't only just to characterize this but also in the new diagram the hypothesis was this other person should be because All these properties and all of these properties that in fact expressed in a language that they've just learned so to me it was All these doubts about whether people could learn kind of went away after I've seen and we had about Quite a number of students, but eight different homeworks and all of them were of this sort Now you don't know one necessarily to spend time manually doing things So when we have large graphs, we are thinking how to automate a better Readability of the graphs. So if you have a graph like this, we would like to separate things into different boxes So you can see groups and sub communities in the different boxes. So there's lots of research going on in different layouts So this is the previous one that does the clustering into groups and the second one is actually moving the clusters around Okay, now I would just like to conclude I have a couple more minutes just to conclude Reflect a bit of what what it all means and why this is so important. But first of all we are thinking about Computing technology and how it's mediating Communication and that has lots of implications for everybody in By the you know for the government's thinking about how to enable the society and how it can influence the society through the media and Furthermore, we know that It is absolutely impossible now to To imagine any big event without Twitter. You know, this is this is this was actually four years ago Almost now when Obama had his campaign on one side. He's talking to the crowd on the other side There's even more crowd talking about him. So this symbiosis of a technology and if In our normal kind of physical life practices are coming so strongly in a very very close symbiotic relationship This is colleague Ben Schneiderman that I mentioned from University of Maryland. He and colleagues at the University of Maryland are working Very closely with the government's understanding how they can increase the social participation of individuals and that applies everywhere Wherever there is a need to help each other humans can help each other How can we connect them and then inform them and motivate them to act? So there's lots lots of activities going in that in that place. Yeah, but one thing this was very intriguing to me Just as you may remember Yeah, it was during the Roosevelt time when Einstein send an email, sorry send a message to a Roosevelt About about the the issues about the nuclear power that just appeared So they're always positive and negative things whenever there is a there is a big discovery or a big Change either from technology point of view or scientific or research point of view One has to be very be very careful and what has happened with the social Media, this is exactly what we are now observing the same thing I think Obama got a letter from Ben Schneider and couple of colleagues basically informing him that social power has arrived and Many different ways to think about it and many different ways to exploit it Therefore we need to do lots of research and understanding because this time we are directly working with human lives Now a small contribution that we are trying to make is based on another Excel Activities we have started a social media research foundation Which is a non-for-profit organization to promote education and social network analysis And so if you are interested, this is a portal where you can find more more information About social media network analysis those people who are enthusiasts and would like to use software and start Creating your own networks. You can you're definitely not going to be alone We created a portal where you can upload your graphs and you can share Information about your data So this is a quite an active community of individuals who are sharing experiences how to create and the graphs how to analyze them and how to interpret them For example, if you click on any of these Images you will get a detailed graph now What's also very nice if you're really proud of your Layout because you can model this you can change the colors you can change the size It's really become extremely kind of personal expression You can you can save this and other people can use it and you may guess five stars. Who knows? Finally if you are interested in the resources on a network analysis based on our experience we've kind of put together a Book it's a little bit outdated now because it's about two years ago And it really explains the basics of the network analysis and it shows how the this spreadsheet paradigm can be used to analyze data and Reinforce this notion that yes indeed social network have been created because of the digital and computing technologies But it's absolutely critical to enable domain experts who are not necessarily computer scientists But domain experts to be able to use this data and gain insights Thank you very much