 And so I'm going to talk a little bit about making pictures with data. I also think of this as playing with data, and maybe you'll see why in a little bit. I feel a little awkward today because I have a great big band-aid on my forehead. And I was going to tell a joke about band-aids, but I figured it would be hard to pull off. Okay, so visual analytics. So I want to talk a little bit briefly about what I mean when I say visual analytics. And then I'm going to talk about what data visualization is and how those two things are similar and different. And then I'm going to give a data demo. Now, I can't actually give a demo today. I was going to give a demo and a program called Tableau to show you what it's like to play with data. So I have some screenshots and some pictures and I'll talk you through it. And then hopefully if I'm good, I'll have plenty of time for questions. And hopefully you'll have plenty of questions. And so let's see. Where do I have to point this to make it work? It worked once. Okay, next slide. Look, there's a little line here. Okay, good. We're back. So visual analytics is the science of analytical reasoning facilitated by interactive displays. Visual analytics combines research from a whole bunch of different fields. And really probably a lot of people say that it got its big funded start around the disaster in 911 when the World Trade Center fell because computer scientists and visual cognitive scientists and people that study data and think about how to analyze things like that came together and they said, well, we need all these things together. So visual analytics is a lot about computers and humans together. Right so it's not just computers it's not humans it's a place that intersects those two things, mainly to make things interactive. So that's actually one of the fundamental differences between what I would say is data visualization and visual analytics is that it's those two things working together. Next slide, interactive exploration. So, and what we mean when we say exploration is sometimes you have so much data. I know there's interesting stuff in there, but you don't know what even questions to ask about it right so if they're looking at a bunch of data to look at something like 911 what happened then. They had so much data they just couldn't figure out how to use it all how to make sense of it all so visual analytics is intended in part to explore data. It's useful for developing insights and questions so if you've got a business case you've got some sort of big business problem you've got a lot of data you're, you're sifting through it you're trying to find something interesting in it. That's what visual analytics is for. Okay, so that's visual analytics. These slides go very slowly this is more like data visualization. Data visualization is less about being interactive. It's more about static pictures visualizations of data. These are all just visualization examples to get the point across. One of the great things about data visualization is, we can use pictures of data to tell a story to make a point to communicate. When you get a data science degree and you go work for a company and you found some interesting things in the data, and you want to convince upper management that they shouldn't go by that other company, where they should go buy these other things. Right. You want to come up with pictures that tell the story quickly effectively so that they can make decisions about them. I don't want to actually imply that there's a real strong separation between data visualization and visual analytics. The point here is just that we have two classes when it's going to start for the first time. This coming fall that's a visual analytics class, and we already have a data visualization class called information visualization on the books. And information visualization is more like this stuff slide. So here's just a chart that kind of contrasts those two things, but the difference or the distinction between the two is a little bit fuzzy. So in our new class, we're focused on visual analytics, we're using tools like Tableau and Watson analytics to make interactive systems that enable you to play with massive data sets. And the class we already have on the books is data visualization actually called information visualization. And it uses it teaches how to use a programming language like are to make static visualizations, and then how to use a design program like Adobe Illustrator to make it beautiful, so that it helps tell the story. Right. If it's designed well, then people will engage with it. That's the idea next slide. So the story I want to tell you today is drawn from a data set from some platform, you might have heard of it it's called Facebook. And Facebook allows people to buy advertising that targets specific audiences. The data that we have is actually from the 2020 election. And we're focused on Biden and Trump but we actually have the data from all the candidates that we're running before the primaries finished. When I expand all this data out. It's a reasonably good size data set it's 41 million rows, it takes up about three and a half gigabytes on my hard drive. And the, the team of researchers here using this data set includes Jenny Stromer galley, Brian McCurn and Patricia Rossini is in London someplace, and myself, and if you're more interested in the illuminating project, you can check us out on the website. Next slide. So this is Tableau. So I was going to demo this so this is the slide version, and you can't see a lot of the text but that's okay. Right. The text isn't super important here. What's important here are these bars Jay can you come up here for a second. So I'm asking a friend to come up to stand next to me for a minute. And I want you to notice something. Which one of us is taller. Don't answer that you knew you knew as soon as he came up here. Humans make comparisons, and we do it so blindingly fast that we do it before we even realize we've made comparisons and visual analytics and data visualization, take advantage of the fact that humans make comparisons. Humans find patterns very very easily, and we make comparison. Okay so in these bars. Oh wait, go back. In the bars. There's a couple things I want to point out so there's a set of bars on the top. That's Joseph Biden's spending for different age groups. And the bars on the bottom that's Donald Trump's spending for different age groups and the age groups. The first shortest bar is 18 to 24 the next bar is 24 to 34 so on and so forth until the last bar is only that was spent on ads for people over 65. And we can look at these and instantly get some insight. One of them is Trump spent more money targeting ads on Facebook. And the other is, they targeted different age groups. Right. And you, like I didn't even have to say that out loud you already know that. But I'll say out loud that Trump typically targeted older people. And Biden typically targeted younger people. Next slide. Now this is Tableau. And what's cool about Tableau is, you can actually just drop in drag and drop your variables, and it will instantly update the picture. And even though I have 41 million rows in here, which, if you're not very familiar with it, that's a pretty good size data set. Would you say it's a good, it's a good size data set. So, so I just dragged this and dropping it updates and creates the picture quickly. The reason why that's important, not because I'm, I'm not trying to brag that I have great big data. It's that the thing about visual analytics is that it's interactive and if the system slow, then it doesn't feel as interactive. It's what gives you a sense of intimacy with the data that helps you find insight. Okay, now this data is a little more complex. I've still got Biden on the top and Trump on the bottom. The top set of bars that are a light blue. That's what was spent on women, and then the dark blue is men. So I've got Biden's women and men and Trump's women and men. Now we did some of the stuff that that Jeff salt was talking about with this data. We did something called automated classification, where we coded a bunch of a bunch of these ads for what kind of message it was. And then we ran algorithms to tag all the data so that we knew what every single kind of add what we've got three types. We've got ads that just try to engage with people. We've got ads that try to ask people for money. And we've got ads that encourage people to go out and vote. And that pretty consistently covers most of the things that that our political candidates are doing with ads on Facebook. Okay. Now again, we can look at these pictures and pretty quickly get some insight. So for example, I would say Biden spend a lot more on women than men. I would say Trump is pretty close to about the same. And Trump is targeting men kind of more evenly but he's targeting older men more than younger men. Biden is kind of the opposite he's chart he's targeting more younger people than older people. And he's really trying to engage with women. What else can I tell you, I can tell you that. Yeah, by that Trump's got this more. He's much more focused on older people. And the other thing I'd say about this is. Yeah, that's enough. This comes up really fast. And it's important that it comes up fast like I said, okay, next slide. Now so the other data that we get from Facebook is where these ads are targeted. Okay, so we have multiple dimensions right we've got where it's targeted. We've got whether it's male and female so we've got gender, and we've got age. In this particular one I've got maps of the mind of advertisement, whether it was engagement fundraising or voting for which candidates. It's darker green if it's more spending and it's lighter green if it's less spending. Okay, so one thing I noticed right off the bat, Trump is on the bottom, and the darkest state in all three pictures is Florida. Does anybody remember the 2020 election. Okay, so Trump did win Florida. Interestingly, Biden's got a more even distribution of how he spent spending money, but he is still spending pretty heavy in Florida. Biden is covering Pennsylvania, North Carolina, Georgia, Arizona, Nevada, Michigan, pretty heavily in addition to Florida, what states did he win. I think all of those, did he win North Carolina he didn't win that one right. So he got pretty good. He, he had a different kind of a strategy. Now Facebook isn't the only place that they're advertising. So these numbers are just about, you know, how he's engaging on social media platform. And there's some other insights that we can get from this. If we were in an interactive environment, we could draw, we could roll the mouse over each state, and it would give us more detail that's called drilling down and drilling down is important because it helps us get detail. And in fact, we could hook up ads to an interface, and you could drill down you could click on Florida, it would bring up the collection of ads that that showed up in Florida, and you could drill down and show me the ads that Biden targeted for women between the ages of 24 and 34 in Florida. And if we had enough detail, we could actually see it per county right so we can really, that's data science right. The high level aggregated view, and the ability to drill down helps us tell stories. In fact, we could click on voting for Pennsylvania for Biden, and voting for Pennsylvania for Trump and we could say how are these ads similar or different. Right and that's, that's the comparison that's what humans do really well. There's some other observations here. There's a difference between fundraising and voting. Biden is asking Californians for money, but he's not asking them to vote you know why. He already knows he's won that state it's California. It's like one of the more liberal states in the country. He doesn't have to tell them to go out and vote of course they're going to vote. So it's a it's a sophisticated strategy Trump's doing actually something similar if you look. Texas is, he's engaging with Texans and asking them for money, but he's not worried about their vote. Right. He's more worried about Florida. In fact, Trump has a very Florida focused strategy throughout. Okay, so we're doing analytics here you know I'm like live streaming analytics I could do this as a podcast and be very fun, but it would miss the visual element. So those are pretty much straightforward visualizations. One of the things that Tableau does is, it allows us to build what's called dashboards and dashboards give you multiple visualizations. You can simultaneously see things happening at the same time. Okay, so I've got my map up there right this is the map that shows what they're spending it's not broken down by engagement and fundraising, but we did see this lower visualization before right. This is gender and age, using the different types of messages that they use. And now we can start to ask some more detailed questions right. If we look at the next slide. If I click on one of these states, Florida, the big green one. It's green incidentally because it was money and I thought okay that makes sense. Okay, so if I click on Florida, the plot actually changed. We can see there's hardly any fundraising from Biden going on in Florida. He's not asking Florida and Floridians for money. He's just engaging with them. And he's trying to get them to go vote. And and Trump on the other hand is asking Floridians for money, but more so he's focused on getting on the vote and engaging with people. And again, we see very strongly by I'm going to engage with older folks strategy of play in Florida. Let's look at another slide. If I click on Florida again it changes the view so if I click on Pennsylvania again it changes the view. Biden's really engaging with voters here. He's asking for a little bit of money, mainly from women, but not much else. And we can see Trump is hardly engaging with Pennsylvania at all, which is funny because you really need to win this state if you're a presidential candidate. He didn't have enough strength in some other places. Next slide. Is anybody in here into politics. So in a way that's great because because when we're exploring data, we can explore data that we're unfamiliar with like I could look at Jeff's credit card data and probably get some insights about the way people are using their credit cards, even if I'm not an expert. I am a domain expert so I'm able to translate these insights into a story that fits with what happened in 2020. Here's Texas, and Texas is a big fundraising state for both people, even though Biden lost there. He was getting a fair amount he was doing a fair amount of advertising to get funding from Texas. And interestingly, neither one of them is voting much. There are some more slides. I'm going ahead and click, click, click. And so the main point to this is is that if you're using a platform like Tableau, then you can sort of interact with the data, and it really does feel like playing because you're kind of learning something at the same time that you're just clicking around. And if you take visual analytics as a as a class. Then we teach you how to use Tableau, but also how to think about what's going on with the data. And we spend time teaching you how humans think right like I don't always just have a tall guy stand next to me. There's other ways that we can talk about how people, how people think and perceive right visual analytics is in part about what we perceive. And then, as a result, you end up with this idea about what's the best visualization to help me tell the story that I want to tell. That's a lot about what it's about. So these are the two classes that were that we're offering. And if you take information visualization, then you learn a lot about our scripting you don't have to be an R scripting expert to take the class at all. It's mainly about static visualizations, and we spend time learning about design elements. It's more coding and less theory is how I would say it. Analytic dashboards uses Tableau instead of our there is a little bit of our or Python in it but not much. It's about dashboards and dashboard design. So while we do talk a little bit about design. It's less about beautifying and more about space. Where do you put these things. It's more about interactive visualization and less about static visualization, and it's more theory and less code. All right, last slide. I think it's the last slide. Yes, questions. Anybody have any questions. Doesn't visualization look fun. Yes, what's your question. I don't know what the data that existed that made the candidates, send money in the area and then what was the outcome of voting. Those three things on top of each other. Yeah, we could totally do that. The question is, because it's in line with my own research, the question is about, you know, can we look at data outside of the data that we got like the polling data that that the campaigns used to make decisions about where to spend money, and then overlay on top of that the voting, the voting parts easy that data is out that data is on Wikipedia so I could just grab it. There's a lot of data there's a lot of polling data out there, but a lot of times campaigns actually use private pollsters. So if you remember from 2020 there was a lot of talk about how the way off. Somebody. Okay. How the polling was way off. And that was in part because they were using public pollsters, and they were not covering the population very well, but the private pollsters were. So they actually had a better perspective of what's going on. Question over here. Yeah, so I have several questions from the chat. How does geospatial data science fit into what was shown during the presentation, if at all. In this case Tableau makes geospatial visualizations really easy. I just used state names. But if you happen to have latitude and longitude for points if your data covers different kinds of geographies. Tableau has got most of that stuff built in it just handles it for you. And it makes it easy, easy is good. Another person asked if we could touch on accessibility in data visualization. Accessibility. Yes, in fact, Tableau has built in color palettes that are colorblind safe, which makes that part of it a little bit easier. And then of course, you know these days people are adding tool tips and stuff like that. So there's ways to take your Tableau dashboards and move them over onto web pages where you'd want to add that kind of accessibility stuff. It's a little harder with the static visualization stuff, but there are actually color palettes and things like that that you could add and then if you were going to use your data visualizations on a web page, you'd want to add the tool tips to the forum. And we had one more from the chat. For the Facebook data, you mentioned a process for coding each part of the topics, which were then used to generate some of the visualization so I think they were just wondering if you could expand on that. We went through the sometimes mind numbing process of looking at hundreds and hundreds, actually thousands of ads, and then determining whether or not these ads are engagement, fundraising or voting. A lot of times it's very easy. You just say, Oh, this is definitely fundraising, right there. They're clearly just asking for money. But sometimes those those the distinction between those become very challenging. And once we come up with something like three or four thousand example ads that we coded ourselves, then we train our models with those and we train them in very similar ways to what Jeff was describing. So we use that gold label data, we hold some of it back as test data, so on and so forth. Any other questions before I leave. One question in the back. What was the most interesting insight that came out of the related project? Well, a lot of it was truly fascinating. And actually, you know, the real thing was how clearly the data showed what pundits have been saying all along, right pundits have been talking about how Donald Trump was focusing on older people. How Democrats are mainly focused on women and younger people. And we just see all the things that you hear about all the time, very clearly echoed in the data. And it was really, I think one of the things that was most surprising to me was Trump's all in on Florida strategy, like why didn't he go for Pennsylvania why didn't he spend money in Arizona. Those things might have made a difference for his election. Thank you for the question.