Last Thursday, my co-instructor and I showed up to the Data Visualization course we teach at the University of Washington with a bag of rocks. The bag consisted of a fairly diverse collection that I myself put together across a set of treks in various regions of California. Our students are fairly used to the quirky, hands-on activities we ask them to participate in most classes, but this seemed a bit out there, even for us. In this article, I’ll focus on the following two points, which collectively speak to the importance of domain-specific integration into data science Education: A description of the actual task we had students do with these rocks. A deep dive into the discussion that followed—which largely focused on the point of making them do this and its deeper connections to data science. What to Do with a Bunch of Rocks? Once the students were seated in their respective groups, we asked them to do the following: Choose two rocks per group. Attempt to formally identify the rocks without the aid of any internet or mobile apps. At this point, most students made it as far as determining if a rock appeared to be igneous, sedimentary, or metamorphic. Refine their initial guesses by now taking advantage of their electronic resources. Students now got much more specific, identifying scoria, slate, red jasper, gneiss, and a host of other rocks in the collection. Design and implement a chart (using software or on paper) that either compared the qualities of their rocks or displayed engaging information about one of them. They were encouraged to search online for supporting data, such as hardness, mineral makeup, potential uses, and so on. Once finished, they submitted their visualizations to us, and we proceeded with a class discussion. What Do Rocks Have to Do With Data Science? Quite a bit, as it happens. As we went around the room, students shared a host of insights about their various rocks. In many cases, the discussion focused on the utility of a particular visual approach students had taken. For example, one group chose to compare their two rocks via a data table that included various points of relevant information. This led to a discussion on how data tables are in fact a type of data visualization, especially useful in two situations: When you have a limited amount of data When it is important that the user be able to pick out precise pieces of data for their purposes Other conversations revolved around the effectiveness of area as an encoding, the particularities of color scales, and so on. All standard discussions for a data visualization course. Once we finished this initial conversation, I posed a more involved question for the class: “So far, we’ve talked about standard visual elements of a chart. We could have discussed these with any kind of data. So why go to the trouble of bringing a giant bag of rocks to the class and asking you to identify them? What’s the point?” The class stared blankly. The moment dragged. Then, one student hesitantly raised his hand. “Um … so we can get comfortable working with unfamiliar domains, or something like that?” Precisely! We’d mentioned this point sparingly to the students before, but this activity really drives the point home. As eventual designers and engineers working in data visualization—and, more broadly, in data science, it is essential for these students to know how to work with domains they may be unfamiliar with. The same goes for you if you are reading this article. As the data expert on a team, you will rarely also be the domain expert, and you must adjust to the data given to you. Sometimes quite quickly. In a previous article, “The Three Building Blocks of Data Science,” I dove into this point in greater detail. The first two building blocks—statistics and computer science—are incredibly important. That said, the actual data comes from the domain. Without the domain, there would be no need for data science. As a data scientist, while you will have the support of a domain expert, you will still need to design solutions and write code corresponding to data you may be deeply unfamiliar with. As such, it is incredibly important to gain exposure to this reality as part of one’s data science education. My co-instructor and I teach in a design and engineering department, with students largely interested in pursuing fields such as UI/UX research and data engineering. We chose to make them work with rocks precisely because we knew they were unlikely to know too much about them (at least at the level of detail needed) beforehand. And that lack of prior knowledge made all the difference. Final Thoughts If you’re reading this, I’m guessing you’re training to be a data scientist, or interested in doing so. Perhaps you already are one and are just rounding out your knowledge. Whatever your position may be, my point remains the same: Every chance you get, expose yourself to new data. By its very nature, literally every field, every discipline, every topic known to man has some kind of data, and an associated group of people interested in gaining insights about it. And the person they turn to for help might just be you. The post I Teach Data Viz with a Bag of Rocks appeared first on Towards Data Science.