View transcript
- Hello, I'm so glad everyone is here, and I'd like to bring us from the lab back to the Exploratorium, a happy place where we could wait and all work together to reinvent the hammer. So, the Exploratorium is known for as many things to many different people, but one of the things that we are known for is a place where we give visitors direct experience with phenomenon to encourage them to make their own discoveries and to really enable exploration, hence, I guess, the name Exploratorium. So, we've done this for the physical sciences, the biological sciences, and more recently, even for the social sciences, so as you walk around the Exploratorium, I hope you'll get a chance to see plenty of examples of these types of exhibits where visitors are really engaged directly with the phenomenon, whether it's racing differently weighted wheels down a racetrack, or looking at the beating heart of a live chick embryo, or trying to suss out how much you trust your partner not to squirt water in your face. But as Jen mentioned earlier, there is a lot of phenomenon that is not directly assessable. They're too big, they're too small, they're too dangerous, they're too expensive. They work on a time scale that's too short, work on a time scale that is too long, so we can't really give visitors direct access to these phenomenon that are increasingly the focus of much of current science. So, our question was, well, can we at least give them access to the data that characterizes these phenomenon? Can we use visualizations to really encourage exploration in this way? So, to try to answer this set of questions, we embarked on a multi-year project funded by the National Science Foundation and the Moore Foundation called Living Liquid, and Living Liquid more specifically tried to answer the question, can we engage visitors in exploring scientific datasets with interactive visualizations, so can we support them in asking productive questions? So, productive means that it's answerable with the dataset, and can we support them in interpreting the data that's visualized? So, in order to do this, we did iterative prototyping for three different scientific datasets. I'm actually going to pull most of my examples from the first dataset, which came from the Darwin Project, which was really all about the global distribution of plankton in the world's oceans. We did do this for three very different datasets because they were so different, so it kinda helped us come up with some more generalizable findings, but they all shared a common characteristic in that it was really all about life in the oceans, hence the name Living Liquid. Thank you, Jen, for that name. So, we prototyped visualizations for each one of the datasets, and we wanted these visualizations to be multi-user because most visitors who come to museums come with other people. We wanted it to be interactive, and in fact, in the very early stages of the project, we experimented with using tangible user interfaces. We needed these exhibits to be standalone because staff facilitation is really not the typical experience that visitors have at our exhibits, and we wanted it to work for not just the governor of California, that's Gavin Newsom right there, but for any casual visitor, eight years old or older. So, it's really hard to summarize a six-plus multi-year project in 15 minutes, so I'm going to just give you six lessons learned, and these are really from the many, many prototypes and evaluations that we conducted as part of the Living Liquid project. Lesson one, it takes a village, or rather, it takes an interdisciplinary team of research scientists, computer scientists with expertise in visualization, and museum professionals. So, the research scientists, they brought the data. They knew the data backwards and forwards, and they really taught the rest of the team how to understand the data to be visualized. We worked with VIDI from UC Davis, and they brought the computer science visualization expertise, how do you encode the data, how do you actually implement this and make this a reality? And I guess the Exploratorium brought the noise and the funk. We brought our expertise and what actually happens in a museum and what might actually work for our visitors. And I would actually say that this collaboration was one of the most useful mechanisms for resolving some of the tensions that come up during design, especially the tension between scientific accuracy and making something understandable for all visitors. Lesson number two, curate the data for exploration. So, this is just one figure from a research paper done by the Darwin Project, and the Darwin Project dataset have 48 different plankton types. Every one of these plankton types have multiple and its own metabolic and physiological parameters, and they were working with at least four different environmental factors varying over time, so not everything could visualized. Their visualization is a very, very thin slice of that very, very large and complicated dataset. So, how do we visualize this for visitors? Well, one way we did it was we looked at the data, and we tried to identify productive questions that were meaningful for visitors, so went out onto the floor, and we talked to visitors. We asked them, what do you wanna know from scientists who study this kind of thing? When we had visualizations, we would talk to the visitors. What kind of questions come to mind when you look at this? What are you curious about? And we really tried to map those questions to what, is this a question that can be answerable with the data, and what would an answer even look like? There are questions that visitors asked that really didn't have, you couldn't really answer it by examining the data, but we found them really useful in trying to understand what kind of context setting do we have to give for visitors to start make sense of a visualization. We listened really hard for how people talked about the content and the data, and we tried to pick out what's familiar to them because if we don't have anything familiar with the dataset that visitors can hang onto, just figuring out what the visualization is about becomes incredibly difficult for them, and let alone making it for us, making it motivating for them to even look at the visualization, do data exploration. So, for the Darwin Project, guess what we identified plankton, and four different plankton types for three different environmental variables, and the questions were quite simple. Where are they, when are they there, how are they different, why are they there? Lesson number three, give immediate access to the data. You might think this is obvious, but at the very beginning of our work, we actually, see if I can start this. This is a movie I think Jen showed earlier. We actually started with the movie of ocean currents, and the thought was this would be appealing, it would get people kind of excited to dive into the ocean and look at the light forms in the ocean, and we also thought that it would give people a little more context for their exploration. But guess what, this was so mesmerizing that no one cared about the plankton anymore. So, why don't we just visualize the plankton, and that's what that is over there, the four different plankton types as represented by the four different colors. Use interactivity to direct the exploration. People really, really focus on the interactive elements, and this was true for all of the prototypes that we worked on. It really worked best when the interaction maps to a task that you perform as part of data exploration, like searching, like filtering, like sorting, so in the case of plankton population, it is these lenses are the interactive elements, and you can move them around, and they essentially allow you to search through the larger database and to filter on the local conditions. Layer in the richness, so it's still a really complicated dataset, even though we took a really thin slice of the overall research data, and we were greedy, and we wanted visitors to have this rich experience where they could really dive into the data, they could explore the environmental factors and start to really answer some of the harder questions, so not just what is there, but why is it there? So to manage complexity, what we did for plankton population, which is based on the Darwin Project, people don't get to see the second level data. They don't get to see the environmental conditions until after they've played around with the rings, and they had to do a little side tap on the ring, so it's kind of hidden from immediate view and the exploration. In this way, we were hoping that we could kinda manage the complexity of the data. And for my last one, data exploration can happen. It's hard to do, but it can happen, and these are just some of the quotes taken from a think aloud study that we conducted with visitors we recruited to look at plankton population, and they did ask and answer their own questions. I think all of the groups that we recruited asked at least one productive question, and almost all of the groups saw some sort of pattern in the data about the plankton. So, data exploration can happen, but it can take time, so each one of these purple boxes is a data interpretation comment, and they're plotted according to time. So, this is great. However, the average time it took for a group to get to their first data interpretation comment was 43 seconds. 43 seconds is a really long time in the exhibit world. In fact, in a separate study, we found that people, when they're not recruited to use the exhibits, spend around 47 seconds in total at the exhibit, so we asked, what are they doing before their first data interpretation comment? They're doing orange things. They're doing decoding. So, they are just trying to figure out what are these purple things doing, what's this shape supposed to mean? So, what I think this is all saying is that data exploration, we can support it, but it depends on creating visualizations that visitors can easily decipher. So, that was just some. The original set was a set of 20. The thing about this kind of work is that the devil is all in the details, so my hope is that throughout this conference, we'll all get a chance to talk about the details, and we'll get also a chance to pop up and really think about common challenges we're all facing and share some of the useful lessons learned. Thank you.