Now that you've learned how to describe and deconstruct every visualization into its major units, there is an important question. The question is, how do you actually know if the visual encoding that is used in this visualization is good, right? Ultimately, the reason why you are learning how to read, and deconstruct, and decode visualization is because we want this step to be useful for something. And as we said before, it's useful for evaluating visualizations and for designing or redesigning visualizations. Okay. So now, I want to introduce two important principles that can be used as guidelines to understand or reason whether a given encoding is effective or not. So the two principles are the principles of expressiveness, and effectiveness. Let me start with the expressiveness principle. The expressiveness principle states that "the visual representation should represent all and only the relationships that exist in the data." What does it mean? It means that the visual representation should represent the information that is actually present in the data, but even more important, it shouldn't convey information that is actually not contained in the data. More specifically, two examples of where this might happen is when ordered data appears as unordered, and the other way around, unordered data appears as ordered. So, let me walk you through a few examples that show you how the expressiveness principle can be broken, okay? The first one is sort of line chart. This is a line chart where on the x-axis, instead of having ordinal or quantitative attributes, we have a categorical attribute, a number of categories. So that's a problem. Why is it a problem? It's a problem because when we look at the line chart, we perceive a line chart as some continuous information that is going from left to right. Right? And in addition to that, most of the time we perceive this as being something that changes over time. Probably because line charts are used very, very frequently to visualize temporal data. Okay? So this doesn't work. It doesn't work because when you see this chart, you perceive a line chart as something that is ordered. But there's no order here. Since on the x-axis we have a number of categories, these categories can be reordered in any way you like. And by the way, when you change the order of these categories, you also produce different patterns of the line chart, and these patterns are not meaningful. So, that's an example of breaking the expressiveness principle because the chart shows information that is actually not contained in the data. Let me give you another example. So, this is a heat map, a matrix visualization, where we have a categorical data on the y-axis, categorical data on the x-axis, and the matrix produces a number of cells, and within each cell we have a quantity represented with color intensity and color hue, in this case. Okay? So, what you see here is what is called typically a diverging color map. Look at the color map at the top right. So, in a diverging color map, we have two color hues, in this case blue and red, but we also have intensity. Okay? It gets darker so it's brighter in the middle, and gets darker as you move towards the ends on the right and on the left. So, what is the problem here? The problem is that a divergent color map is used when it makes sense. And when does it make sense? When there is zero value or there is an average value that is in the middle and we want to emphasize the idea that there are positive values in one direction and there are negative values in the other direction. But in this specific case, that's not what is shown here. This is just that continuous sequential attribute that is mapped to color. Okay? And there's no middle value, there's no zero value, but when you look at this heat map, at this representation, you perceive it as if there is zero value and some positive values that are the blue ones and negative values as the red ones. But again, this is an example of breaking the expressiveness principle. Why? Because the visual representation communicates something that is actually not present in the data. Last example, is a bubble chart where bubbles represent different instances in the data set. The size represents some quantity, but color and position don't encode any information and that's a problem. It's a problem because when you look at something like this, you try to decode out of it a meaning of color, and probably also a meaning of position. But in this case color doesn't encode any information, and position doesn't encode any information. So again, once again, this is breaking the expressiveness principle. Why? Because the chart, the visualization is communicating information that doesn't exist in the data.
