How Data Visualization is Being Used to Mislead You (7 Misleading Tricks)

1. What We Will Cover: Hi there. My name is Joshua Brinley On In this cost, we're gonna be taking a look at the ways in which misinformation is spread online, using data that isn't actually inaccurate. They're actually presenting the rial values and the figures, but that torturing those numbers and twisting them in a way that is miss informative to deliberately spread misinformation. So what we're actually gonna be covering in this cost is how data is used to spread. Misinformation will be looking at some of the common techniques and tools used to torture accurate data to twist it and turn it into an inaccurate message will also be going through some real life examples off graphs that have been posted online with the deliberate intention to mislead on also how you can spot these types of graphs on being more aware of them. You'll also learn how to avoid creating such misrepresentative graphs in your own data visualization. If you're into data signs or your business analyst or use data to present information, did you know that over 2.6 quadrillion bytes of data is produced every single day? Data visualization is the method we use to communicate that data graphs on charts online are all over the place reflects on all of the graphs that you've seen recently online on news sites on TV, Social Media Advertising's They're all using data visualization as a convenient way to spread the actual data. Well, unfortunately, in many cases, the intention of these graphs is actually to misinform. It's not to create an awareness in their audience of how the data should be. They're actually trying to misinform you by misrepresenting that data using tricky techniques in visual design. So we're gonna start exploring these. So join me in the next lesson where we start exploring how accurate data can be presented accurately, but in a way that misrepresents what that data is actually trying to say, I'll see there. 2. How Data Spreads Misinformation: either and thank you for continuing of the costs in this session. We're going to take a look how data could be misrepresented while still remaining accurate . It's easy to tell a lie with numbers, but impossible to tell the truth without them. So this is an expression from undressed uncles on. I think it really accurately summarises the issue we have with numbers. You may, as someone who creates data visualization, have heard someone in business say something along the lines off. I don't want the numbers. I want the truth. But what does it mean? Numbers are cold. They're statistical, that one biased. So how can they not be truthful? Well, if you manipulate the way he visualize and present those numbers, they can be entirely on truthful, so numbers a cold, unbiased fact. So how can they mislead? Well, there's actually plenty of tools and techniques used to display numbers so that they say something that isn't truthful, even though it is accurate off. Fortunately, we live in a world where politicians biased new sauces and companies are intentionally presenting numbers in a way that misleads without outright lying. Some of the places you come across misleading data is places such as the news, where partisan issues with a biased slant presented as neutral. Even though the numbers themselves are presented neutrally, they've manipulated the way they've presented those numbers to lean one way or the other. Social media is another place, and it's possibly the most insidious way in which data is used online to misrepresent. Often companies or political parties share graphs online on social media. You might be scrolling for your Facebook feed, and you come across a simple fact or a small graph and think nothing of it. But this graph was placed there intentionally. Often you can't attribute the graph to an offer, but someone created it, just trying to spread misinformation online. And finally, oftentimes, misleading graphs have done accidentally poor formatting and design choices that have gone into graphs accidentally misrepresent the data. So throughout the costs were going to be taking a look at some of the techniques used to misrepresent that data as well as looking through some real life examples. So join me in the next lesson where we start breaking down how these techniques are used to misrepresent the data, how to categorize them as well as looking through some real life examples off graphs that have been posted online with the deliberate intention to mislead. Thank you. I'll see that. 3. How 3D Graphs Introduce Error: hi there, and welcome back to the cost. In this lesson, we're gonna be taking a look at some of the common ways and techniques in which data is represented online with the deliberate intent to mislead. So the common techniques can fall into two categories. The shape of the data elements such as the bars and the lines, or manipulating the graph itself. So these are the techniques used to lie while still telling the truth. Firstly, you have the three d perspective. So using three D perspective, excuse the values of the graph. Secondly, we have this scaling. So scaling two D shapes and how it exaggerates the differences. Let's explore how three D perspective misrepresents data. So his a common three d graph. I don't want you to just take a moment and have a look through it. What we have is the revenue generated for six different stars. Now, at first glance, this may just be a normal looking graph, but I have created this graph in in a deliberate attempt to mislead all the values are accurate, but I've created it in a three d graphing tool so that the margin of error is exaggerated. So what I mean is let's take a look at star for and I want I'm gonna ask you a question. What is the value of stall fall? So here's the graph. Take a moment on. Just measure the value of star for the problem with three D graphs is theme margin of error that they introduce. You can imagine there's actually many different ways in which people could accurately be measuring stuff, for you can imagine an audience of varying, different people on all of them are looking at stuff are, and it's not unreasonable to assume that many of them are getting different values from each other. The problem arises in How exactly do you measure the three D each graph? Let's take a look Imagine. It's a three d bar, as is represented here. Do you measure it from the top of the front facing plane like this? If so, begin a value of about 4400. Well, perhaps that isn't the way you're supposed to measure it. Maybe you measure it from the absolute top most point. So the back corner Well, if we do that, then we come up with a value of about 808,900 somewhere around there? Well, maybe both of these air inaccurate. Perhaps you're supposed to imagine there's an invisible wall sitting behind the bar, and you follow that wall to the corner than measure along the height to the axis. Well, let's take a look at what we get when we do that. Hey, we're getting an exact 5000 so you can see the problem beginning to arise. There's a margin of error between 4400 to 5000 so there's a 600 difference where different people different audience members, particularly those in experience that reading graphs may read this graph in a lot of different ways. So a graph like this is truthful, but it is misleading to a lot of people. We only have six stars here, and we only looked at one. You can actually imagine how misleading this graph would be between all six of those stars . Every single person viewing this graph in the audience will come to different conclusions about what the values of each of those stars mean. So here we have the same graph represented in two dimensions. I asked again. What is the value of star for STO? Far is actually sitting just above the 5000 line. We can clearly see that. So the value is 5150. We can very quickly see that from the graph. No longer are we assuming that it's actually 5000. There's no way it's 4800 and it's absolutely impossible that it's gonna be 4400. You can see there is no room for error in the two D. Graf, Where's in the three D representation of this data? Everyone is gonna have a slightly different interpretation of it. So both grafts accurately represent the data. But one of them did so in a misleading way that can tell a different story. Now, in my example, I only change the graph a few degrees. The three d element of the graph was only at a slight angle. You confirm exaggerate this difference by making him or extreme angle Freedy. And this is what a lot of people are doing online will take a look in elated lesson on a particularly egregious example of three D and how misrepresented it can be. But I hope in this lesson, I've demonstrated how misrepresentative three D graphs can be. There's just no reason to use one. So if you're creating graphs online, don't use three D. And if you come across a glass online using three D, take a step back and wonder is the author trying to mislead if this was it, just a poor design choice. So thank you for joining me, this US, and join me in the next lesson where we take a look at two D scaling on how misrepresentative two dimensional icons can be when used as graphs. 4. Scaling Shapes: hi there. And this lesson, we're gonna take a look. A to D. Scaling on lend house. Scaling images in two dimensions exaggerates the difference between them. So here we've got three different shapes. A circle, a square triangle on. What I've done is I've scaled them up three times, so each of them has a scale of three. The thing is scaling on a bar such as in a bar. Chuy's easy to measure because it scales very easily. It only goes lengthwise in one dimension, but over shape scale both vertically andi horizontally, which means just scaling the volume, not necessarily just the height. So something that is three times bigger becomes nine times the volume. This starts misrepresenting the data. So take a look at this real life example. This is a type of graph that can be attributed to an author. Rarely, it's just saw Stas McDonald's. It's the sort of graph posted online in social media that most people would just scroll past without really absorbing what's going in on it. But it does mislead. It's done so to try and exaggerate how much bigger McDonald's is to the rest of these fast food chains, So what this graph is showing you is the revenue generated for a number of chains of fast food. We have Starbucks, Taco Bell, pizza, harp and so on. So as you increase in size, what's also happening here is their increasing in scale. Take a look at Burger King and then take a look at McDonald's. The difference is massive. On the face of it, it looks to be about 12 possibly 15 times bigger because you can fit the Burger King logo possibly around 15 times in the McDonald's logo. But if you take, look closely at the actual values. Burger King represents 11.3 billion. Where's McDonald's represents? 41 billion. So Burger King isn't even four times the size. But because this shape has been scaled in two dimensions, it implies that McDonald's is a lot bigger than it actually is. What it does is massively exaggerate the difference so you can see McDonald's appears to be about 15 times bigger, whereas in the reality it's about 3.5 times bigger. They've also done something else a little tricky here. What they've done is put the GDP of Afghanistan in there, so this is just one country. This graph could have picked any country wanted, but it particularly an intentionally pictograph that is bigger than Burger King but not his biggest McDonald's. What it subtly implying is the rest of these chains on bigger than some countries. Where's McDonald's represents the value of an entire country? The reality is there are plenty of countries a lot larger than McDonald's. There are also countries smaller than Burger King, but they've intentionally only chosen one country that sits nicely between them, and it's also a visual shape that's been scaled. So McDonald's has been exaggerated to a pay a lot larger than it actually is. Type of graph like this goes online, goes on Facebook Get shared around is perhaps an interesting nugget of information. But what is doing is implying and misleading so that people think of McDonald's as a lot bigger than any of these other chains as it's the place to be. That's the intention of a graph like this, and it's intentionally designed this way, trying mislead its audience. So that's the problem with scaling two D shapes. When you scale in two dimensions, you scale by volume, so something that is only three times bigger becomes nothing times the actual size when used as icons on a graph such as in a kind of bar chop bar charts actually, only scaling one dimension by the height, but because you're using logos or images are picked totals, you're actually scaling in two dimensions, so this has the knock on effect of exaggerating how much bigger things up. So if you're creating graphs online or sharing grass with an audience, try to avoid using shapes. Stick to the common bar. If you come across a graph like this, you can almost guarantee that the offer is trying to exaggerate the difference. The worst part about this is even after you know that the difference has been exaggerated. If you then ask an audience member who viewed a graph like this in a couple days, how much bigger McDonald's waas they'll still exaggerated even after you told them. The graph itself has been exaggerated, and McDonnell it is only 3.5 times bigger than Burger King. They'll still it place it a lot bigger than Burger King because the graph is done, its job. It's misrepresented that data, so join me in the next lesson where we'll take a look into some of the ways in which the graph itself can be manipulated to misrepresent the data that is trying to display. Thank you. I'll see that. 5. Truncated Y Axis: Hi there. In this lesson, we're gonna be continuing on with our misrepresentative graphs. We're now going to be looking at the second way in which grafts could be missed. Representative. Okay, there's four ways in which you can manipulate the graph so that it misrepresents the data that is trying to present. Firstly, you have truncated y axis, which means to start the access from a value other than zero. Next, you have zero scaling, which means the bars don't actually represent a value. Then you have stretched access. So by squashing or even stretching either access, you can exaggerate or diminish trends. Finally, if the data doesn't online to your story, simply omit it. So let's take a look at the truncated y axis. This is when you don't stop the access from zero. It's actually one of the most common techniques used to mislead with data online. What it does is exaggerate small differences between values to seem a lot larger than they actually are. So let's take a look at the truncated Y axis. This is where you don't start the vertical access from zero. It's actually one of the most common techniques used online to mislead and misrepresent data. What it does is it exaggerate small differences between values so that they appear or they seem to be a lot larger than they actually are. So let's take a look at an example of a truncated Y axis on the left. You have a graph that shows the vote between leave and remain jobbing. Brexit, it's asking, should Britain leave you and then showing you the percentage of voters who voted leave on the percentage of voters who voted for Mame? On first glance, it appears as though the remain vote is huge when compared with the leave vote. But what you may not have noticed initially is that the Y axis has been truncated. The Y axis begins at 36 on only goes up to 50 instead of starting at zero on going up to 100. What this does is half the effect of exaggerating how much larger the remain voters when compared with the leave vote. If you take a look at the graph on the right, which shows the exact same data on this time, the access has started from zero and goes up to 100 as a percentage should. Then suddenly That difference doesn't seem so extreme, in fact, that the remain vote is only slightly above the leave vote, which is a much more accurate representation of this data. So a graph like this could be produced by leaning new salts or just accidentally and shared online on what it does is influences voters by spreading misinformation. It's misrepresenting the truth of the matter, even though technically, there's nothing in that cured about the graph. But it's misrepresenting the presentation of those numbers. So let's take a look at this second example. This is the type of graph produced by KFC and shed online in social media. It's not part of in an for It's never gonna be on the news. It's not really worth talking about, but it does spread misinformation. If you look along the bottom, it's showing the calories off different meal choices, but they don't start from zero. They start from 600. What this does is it exaggerates. The difference between the KFC item on the over items was doing is implying that KFC is healthier, the knees over choices. This is this hyper graph that get shared online without much thought from the viewer. They don't deep dive into a study on these things, but it does plant the idea in the back of their mind that KFC could be a slightly healthier option. So next time they find themselves in the food court, they might lean towards KFC, thinking that taking a slightly better option the over fast food choices. This is how this type of misinformation is spread. There's nothing inaccurate about a graph like this, but there's a few different things going on. Firstly, the truncated waxes exaggerating the difference so that KFC appears smaller, therefore healthier. Secondly, it's omitting data. What over meal choices have they picked from these restaurants? You don't know. We'll be looking into admitting data further in a novel lesson. But it's just to point out that graphs can actually use multiple techniques to try missing . Vom we should take away from this graph is that it's this type of graph gets spread online and shared without a whole lot of thought. But it does plan the idea into people's had through use of communication off misrepresentative data so that truncated waxes. You can see how by starting a graph, not from zero you actually exaggerate those differences? Some cases you want to exaggerate how much larger something is. But in the cases like the KFC example, you might be trying to exaggerate how much smaller something is compared to over choices. This also shows and demonstrates how data can be used to misrepresent and spread disinformation online. There's nothing inaccurate about this graph, and it's not gonna be part of a big news site. It's just posted online on Instagram on Facebook. People see it. They don't think to thoughts about it. But it does plan the idea in the back of their mind, an idea that has been curated by a company with a motive to try and get you to buy their products. The fact that it's data seems to have some sort of authority, so it's actually a really effective marketing tactic. Data and Graphs holds some authority as being unbiased representation Ziv statistical fact where, as you can see by presenting them in such a way, they're actually telling an entirely inaccurate but truthful story. So join me in the next lesson where we explore how removing the scale from grass creates misrepresentative graphs. Thank you. I'll see there 6. Removing the Scale: Hi there. In this lesson, we're gonna be exploring how removing the scale from your graphs allows the offer to tell pretty much any story that they want to. So removing the scale means that the bars in a Biograph actually have no representation of what the value of that bar is supposed to represent. It basically gives the author complete control over what it is that they want to say. It's also can be quite difficult to spot. So let's take a look at an example of a bar graph that doesn't have any scaling on the left . We have the Iowa straw poll, a poll taken between use on who they're voting for for the presidential election. What we see is young, his phone away, the winner with 22.5%. And then he's a lot further ahead than Bernie Biden Warren. And so, um, so let's take a look at this more closely. Young. The bar for Young represents 22.5%. Where's the bar for? Bernie represents 21 person, so there's only a difference of 1.5%. But these bars, the bar for young, is a lot longer than the bar for Bernie, but that difference only represents 1.5%. Now take a look at the bar that represents Biden and compare that with the bar that represents Bernie. The difference between these two bars is a lot less than the difference between Young and Bernie, but the difference is 9.5%. So what this means is there's no scaling on these bars. This image is presented as a bar graph. It looks as though it's a Biograph because it's bars presented as though they represent values. But in fact they don't represent value to toll. There's no scaling applied to them. This allows the offer to present Bernie as being far behind young or young, as far ahead of everyone else. If you take a look at the same graph on the right, this is the same data but represented with scaling. This time, the difference between Bernie and Young is very minimal. There's only a slight difference between them on then. The difference between Bernie and Biden is now a lot larger than the difference between younger Bernie, so the graph on the left is intentionally produced and published to try and misrepresent how the political vote is going. Where is they haven't used any scaling a tall on these bods. They've just presented this data as though it's a bar graph. But there's actually no data going on here, so you can see how removing the scaling allows the author to produce any value. They actually one. They could have put Bernie anywhere along that scale on just eggs, completely exaggerated. How far behind young he actually is. They could have said anything they wanted. They could have placed him right next to Young. They could have placed him almost off the page with the difference because they don't have any scaling. So this is how they actually use data to misrepresent the truth of it by presenting it as a bar graph but not actually having any scaling. It's not actually Biograph. It's just a stylistic choice of presenting values that I am that showcase themselves and pretend to be a bar graph. But there's no actual data representation or data visualization going on. So there, intentionally trying to mislead you. So join me the next lesson where we take a look at how stretching the access can manipulate the values that they're trying to represent. 7. Omitting Data: Hi there. In this lesson, we're gonna take a look. A house stretching the access can misrepresent the data, so stretching the axis is when the dimensions of the graph have been adjusted to stretch or shrink the axis. What it does is it removes or exaggerates a trend. It can exaggerate an upward or downward trend or a commute. Remove any variation if stretched enough, for example, a line graph that has been zoomed out enough. It's just a flat line, so let's take a look. That's an example of where the axis has been stretched on the left. What we have is the average annual global temperature presented in Fahrenheit, and you can see it's a pretty flat line. Where is on the right? We have that same data represented, and you can see to clear things from this. Firstly, there's a lot of variation in that line, and secondly, there's an upward trend. So the graph on the left has been stretched. The Y axis starts at a negative value and goes all the way up to 110. They've essentially zoomed out on this data to present it. It's an almost flat line, whereas on the right. They've zoomed in that to that data and properly representative, so you can see that there's a lot of variation on declare slight upward trend. Where's the graph on the left is produced to inform you that there is no increase in temperature and there is no variation because it's essentially now presented as a flat line . So when you zoom out on any sort of line graph, you can present that number to be entirely flat. If you wanted to buy zooming out enough, you could make a trend completely disappear in a line graph. So now let's take a look at the final way in which you can change a graph. To misrepresent data on that is to omit important data points. If the data disagrees, then just remove it. The problem with this is it's really hard to spot because you don't know what you don't know. This type of omission is commonly used in surveys or data over a long period of time or where there is a social factor into the data. Let's take a look at a few examples, so this isn't to start the discussion on Trump or anywhere else It doesn't matter where you sit in a political party, but what this does is show you how data can be used to misrepresent if you simply omit data on the right. We have President Trump's approval rating, so we have an approval rating of 88 cents and a disapproval rating of names them. So by far he's largely approved. But if you take a look at the actual data used to produce this graph, you can see that all these done is chosen the votes that approve him and ignored the votes that don't approve. You can see on the graph on the left. The actual disapproval is far larger than the approval ratings. But if you simply don't show that, then you can show that the data is largely approving. Andi. There is very little disapproval. So this is a case of where data is just omitted to misrepresent, even though there is this little notes on this graph that says it is among Republicans. So there is no lying necessarily going on in this graph. But even after being told this, it's still misrepresent. It's people who aren't Republicans and do disapprove of Trump, but still exaggerate his approval rating after seeing this graph, even though it's been pointed out to them that it's inaccurate, that's the danger of spreading this kind of misinformation. So this kind of a mission of data isn't just in survey results. It's actually quite common technique when using results that contain a social factor or, for example, are over long period of time. In the next lesson will look at a couple of real examples of this, but it's just a note that omission of data can happen in graphs where there's a social factor influencing those numbers. For example, it could show you the number of people dying of a certain disease over a long paid of time . But why? My not factor in is the size of the population there, simply more people. So more people are dying of this disease, whereas the graph itself is just presenting the number of deaths this commis represent because there's a long period of time and there's a a social factor, Teoh, including that those results but the graph, it doesn't represent them, So that is how you misrepresent by omitting data, joining the next lesson where we're going to review some real examples on point out, the different types of ways in which they misrepresent data and combine methods to ultimately tell in an accurate story. Thank you. I'll see that. 8. Real Examples Of Misleading Graphs: Hi there. In this final lesson, we're gonna take a look at some real life examples of where data is, deliberately or accidentally misrepresentative. So let's look at some real examples and understand how dangerous they can actually be. So here's an example of where data is used to communicate in an advert. What it's showing you is the speed of three different Internet Explorer's on the left. We have Google Chrome, which is slower, and it's saying it's slower there. And then in Mac, soft edge we have is the fastest and then Mozilla, 9% slower. So what you notice first is it's using a speedometer, which is a car speed to him to plan the idea that things are very fast. Max Soft Edge is the fastest because it's maxing out that speedometer. Where's they've arbitrarily chosen a speedometer to have the maximum value to coincidentally be the maximum speed of mike Soft edge. But that's just a subtle way, which that communicating speed to you. But the way they're using data is entirely inaccurate. Here there's simply no scale. Mind soft is the fastest, so it's 100% here. Google chrome is only 5% slower, but in terms of the speedometer, it's almost 1/3 slower. And then Mozilla Firefox is 9% slower, but it's actually only 1/3 the speed of Mike soft edge. So there's using data to communicate these speeds. But there's actually no scaling going on here, so they're misrepresenting the data to you to try and sell you this product. So here's a second example, and this one has a couple of different things going on. So what is showing you is the number of murders committed using firearms between the early 19 nineties, all the way to the mid 2000 tens on the state's point represents the number of murders. So what it's kind of showing you is they've planted one sort of indicate on this graph that says, in 2005 Florida enacts its stand Your ground law, which without getting into a huge amount of detail, is basically allowing people to use lethal force if they're threatened or if they're required to. So what this graph is doing is, firstly, it's entirely upside down, so the ground is completely upside down. That scale goes from 873 to 0, so the graph itself first glance appears as though this stand your ground law has reduced gun deaths, but in fact it hasn't. It's actually increased in gun deaths on this graph has been flipped upside down to try and misrepresent that. But secondly, all it does is show you the number of gun deaths. And, as we mentioned earlier, lesson grafts weathers over long period of time, such as this one, their social factors to include, for example, how easy is it to access guns? How in forceful are the police of gun deaths? How better has medicine come along so that when people are wounded with firearms, they have a much higher recovery all these kind of things that will have an influence on the number of gun deaths? But this graph commits all that just shows you the absolute value of gun deaths and then attributes the entire change in that graph to this one law, which disagrees with its point. So it flips the graph upside down. So this is a type of an example where this is not necessarily an accurate data, but it has some manipulation on the graph. It's upside down, and secondly, it doesn't really tell the full story of it because there's a lot of other factors that go into these figures, and this graph simply omits them all. So is our final example. We have a more topical one. So this graph shows you the number of covert depths as compared to a number of other pandemics we've experienced in the past. So this is an example of a graph where the author isn't trying to mislead. It was just some poor design choices. The data is accurate. Firstly, this is accurate. Riel data on the author of this graph presented them in this way through just stylistic design choices, unaware that it's being misrepresentative. And this graph got spread all over the Internet at lots and lots Million's possibly saw this graph, and it had an influence on how they perceived the pandemic. So let's break down some of its inaccuracies. Firstly, everything is using three dimensional Andi circles, so they're scaling in several dimensions, so it's really difficult to tell natural size of certain things. Secondly, it's on a three dimensional plane, so the plane extends backwards into the third dimension, completely skewing how big certain things are. For example, at the back, we have the under nine plague. Where's this? Appears to be a lot smaller than the New World smallpox outbreaks, whereas in fact, is actually bigger because of this three dimensions. The icon that represents them is actually smaller. The antonym plague icon is smaller, but it actually represents a big of value. The New World Small box. This is due to the fact that it's in three dimensions. Secondly, it omits data. It doesn't really show you the population. It doesn't really show you new medicine or any of a number of factors that could influence how a play get spread. All it does is show you the number of deaths normalized toe world population. But it's still omits a large amount of data. So this graph commits kind of every sin we've gone through its three dimensional. It uses scaling shapes. It misrepresents data by omitting certain things on. Overall, it was just created inaccurately, but it was done so accidentally. No one was behind this, trying to mislead. In fact, they were trying to inform, but they made a few accidental design choices that have led to this misrepresentative graph . The offer later made some changes to it, but his revised version was no way near as impactful because it wasn't stylized. The fact that this is so stylized increased the amount it was shared, but it was sharing misinformation. So there you have it. Those are the ways in which data can be misrepresented to try and misinform and spread misinformation online. So I hope you've taken away a few things in this, Firstly, the dangers of spreading misinformation. I hope you're far more aware of how misinformation is spread using factual data online. Secondly, I hope you've been equipped with the correct tools to identify when graphs or possibly trying to mislead you so that you can avoid creating graphs like that under void being misled. We've graphs online. Thank you for taking part in this cost. I hope you learned a few valuable things. I encourage you to share and review and leave a rating for this cost if you liked it so that the students confined it. I'll also leave some links to you so that you can follow me on skill share or over social platforms where I share lessons learning nuggets on just great Andi. But examples of data visualisation. So thank you for taking part in this cause. I hope you learn some interesting things on. I'll see you in my next one. Thank you.