Principles for Better Data Design - Create Context - Vertical

So like our desktops and our glove compartments, your basements are messy. They can be at least, and usually there's only one person who actually understands the system, that being you. A layer of context could be very useful. To illustrate that, let's look at a couple data sets to answer this question. Pause to drink some coffee. If the way in which current events are covered by the news media is labeled as having a negative tone, do we listen to music that reflects that tone? In other words, if the news is bad, are we mad? For the year twenty twenty, the dreaded COVID year, we can see from a dataset of the Billboard top one hundred songs how the most popular songs in the United States trended in terms of emotional labeling. So in other words, green peaks mean that there was more happy lyrics among the most popular songs, and the red peaks mean that there were more angry lyrics among popular songs. So let's add one more dimension to this, which is called the Goldstein metric. This is an expert based measure that scores potential impact from an event on a country's stability on a scale from negative ten to positive ten, which represents highly conflictual to highly cooperative. And if that seems really confusing, that's the point. So let's make it less confusing. First, using analogy and metaphor. So I mentioned this. It's kind of a convoluted description, but they label news events as things that are highly conflictual and therefore, like, might make a country less stable or very cooperative and therefore more peaceful. But it's not immediately intuitive, so there's visual analogies we can draw, for example. In this case, using, like, a very familiar weather app. When the sun is out, we all get along a little bit better. When the chart dips, it gets dark and chilly, and we're more likely to retreat. Right? We're more likely to maybe be a little more pissed at one another. And so visual analogy. Right? Goldstein high, we're getting along. Goldstein low, we're not. For those that are a bit more creative, you could rely on a metaphor. In this case, it's a visual metaphor. You could write one. Imagine if world events were an Olympic sport where cooperation is rewarded with higher scores. So that score is the gold standard. In this case, with this crappy AI generated art, we see some judges judging a military parade, giving it an average score of negative four, which is somewhat of an act of aggression, but not as bad as say a drone strike, which would have to be like a negative nine or something. Second principle within creating context is all about baby steps. So I have this memory, again staying within the year twenty twenty, the COVID year of our lord, dancing, which is strange. To Rain On Me is a Lady Gaga Ariana Grande song. If you haven't heard it, do yourself a favor. It was a very big deal. It came out in May twenty twenty. I was with friends outside, obviously. You had to be. And it was like this brief respite from everything happening in the world. And I kinda had this question, like, I enjoying a rare good news week at the time? Like, were things just like slightly more positive? Or was I like trying to dance the bad feelings, the bad news away? Was I dancing along with the news or in defiance of it is the question. If we look at the Goldstein metric, again US news, from the beginning of twenty twenty to the end, yeah, news got worse. Right? And worse and worse. There are some peaks in there, but it was, like, largely pretty bad. And you can see there are some reasons why coverage became kind of negative. So it turns out that during very negative news weeks, the Goldstein metric was was low, and angry lyrics increased by thirty seven percent. This makes some sense, whereas happy music on average decreased a little bit. This also makes sense. What was less intuitive was this correlation. It's not the strongest correlation in the world, but a correlation nonetheless, where happy lyrics experience an uptick one week after the Goldstein metric goes down. This suggests that there might be some kind of delayed response where new sentiment takes time to influence music preference. Like, maybe I needed seven days to process. Rain on Me was actually released May twenty second of twenty twenty, and it hit number one on the charts. And one week earlier, US COVID deaths surpassed a hundred thousand. Just a correlation, but kind of interesting.