 Welcome, everybody, to the next talk. We're going to be continuing on in the education theme and talking, Cass is going to be talking about lessons learned in education during the COVID craziness. So take it away. Thank you so much. Hey, everyone. So my name is Cass Wilkinson-Soldania. I'm a data science educator at the Children's Hospital of Philadelphia. I use they, them pronouns. And I'm really excited to talk to you today about kind of data learning strategy in complex institutions like many of us work in. And you also hear some specific call outs and also thematic rips from other things that we've heard over the course of the last two days. So this is me. Hi, and this is my cat, Jane, just looking cute. So by training, I am a data librarian and I'm also a mixed methods researcher. So my work kind of blurs together the world's social science, data science, and education. I work with learners who find their way to data science. And sometimes their path to data science is very clear. It's very enthusiastic. Maybe they have their time marked off to work towards this goal. They might have a background to something similar to data science. But honestly, for the vast majority of us, we kind of come into data science maybe a little bit sideways, maybe tentatively. We might wander in a little skeptically or nervously. Maybe we follow a trusted friend or colleague into a particular set of methods. But at some point, we make this hop and we let ourselves be vulnerable. So here at CHOP, I work at the Children's Hospital of Philadelphia, we call it CHOP. I work with this wonderful, diverse group of research data managers, clinicians, analysts, engineers, people who would like to deepen their toolbox of data and open science. And so our program is called ARCIS. And it's part of the Department of Biomedical and Health Informatics. It's an initiative to, as you says here, link clinical and research data to accelerate science. So we're educators, this unit within ARCIS. And our goal is to really support training, guidance, resources around data. And our mandate is this idea almost of culture change, of really trying to help people share their data and feel comfortable. And I really identified earlier what we heard about this idea of the goal of sharing and how that can play out in institutions where there might be interest, but some kind of cultural, logistic, privacy resistance to that. So that's kind of a bit of what we do. And so while we do certainly work with folks who are deep into their machine learning practice and journeys, a lot of times we're working with people who are making some type of hop. I would argue that all of us are making some kind of hop into either the first or the next step in data learning. So for folks who have made it here to this ARC conference, whenever you're watching this, you've likely begun to see that when you have an experience like this, with code like this, it's difficult. It's, of course, but this is also the context where we sort of have learned that this is when we reach out to people. And in fact, whether it's a Google search or a form post or a Slack or talking to somebody in person, these are actually some of the moments where I would argue that we feel the most connected as communities is when we kind of can reach out and say, hey, I have this question. I have this approach. So we're very practiced in dealing with this kind of moment. But for newcomers, this can be very deflating. So I think it's always important to come back to and remember, even now, I like this. This is just flight out console output. Or rather, this is from our console. You can see the output, and you can kind of feel it, right? Still those error messages. And we also are folks who, when we hear that new method, it just gets us excited, right? So I was really excited earlier when we heard about Tranley's tree heater package. My media thought was, oh, I was stuck for so long on this one part of a product in grad school. This would have been perfect. So my mental model was like, oh, interpretable machine learning visualized in a way that you can explain and tease out. Like this fits right into that. But again, for folks without that mental model, it might be harder to integrate some of this information. So as an educator, my task is often to help people, basically, get to the point where they can bridge some of these moments of not knowing and feeling vulnerable to some kind of actual community building around that. So this is myself, when I was working on using the learner package for a project with my colleagues here. And I needed to find a way to test code that could run in both Python and R, where it would run in both, but the result would be different. I knew. I had the strategic cheat sheet of being an educator, of course, but I knew that this was the perfect time to reach out to the Slack community, which I'll talk about in a moment, and get some great, you know, this was a really fun way to reach out. I loved, I think I got to know Paul a little better through this and other folks here at CHOP. And that was a really nice thing. So I'm going to really focus from the rest of my time here on these two questions. So how do we help a learner's leap into data science lead to persistence and belonging? And then also, how do we sustain learning growth in careers? And I'm specific about those things, not just learning, but learning growth in careers within complex institutions. So I'm going to kind of give you some reflections from our experience. And hopefully, it might give you some new ways to think about some of these concepts that probably are important to you in different ways. So one is communities of practice, which we hear about a lot, and is very important. And I will, for a moment, bring us back. I wrote this, and I was like, oh no, bringing us back to February 2020 is kind of tragic, right? That missing in-person instruction is hard. But so this was part of our mode over the winter, kind of a sense of what it meant to be doing data education, data learning within our particular part of the universe of data work of research. So our little ecosystem at CHOP is basically this kind of overlapping communities of practice. And so this was part of ARCIS data education. One of the things that I have done is sort of develop and lead workshops. Like this is a machine learning workshop that will be lighting collaboration with this great group called MetaChop that is kind of cross-institutional. And some of the things that I will draw attention to here is being able to get people together in a room. I love the technique of having small groups. So I think that in Stephens 101, in the 101 session we had on Thursday, there was a similar concept of small affinity groups. I think that kind of thing is really important. I also love giving people coffee and food. So that's something we did here. But this is one modality that we had. And when we work, we also find ways to cross remote and work within other groups. So you see here some of my Slack notifications around my department, DBHI, and then also the CHOPR and the Python user groups. These are amazing, really wonderful groups. A lot of folks here are parts of those groups you've heard from from CHOP. And so they're really gathering points in a context where people can focus intensively on a language and learn kind of what other folks are doing within the hospital with those languages. So this is, you probably are familiar with these different, having a very full Slack slate, right? It's a really wonderful thing. And to give you a sense of what the world looked like in the winter, our cadence was there would probably be maybe about one or two bigger events a month that are in person and then some regular series using R, using statistics. So that was kind of the cadence across these different groups. And so I wanna just draw your attention to when we're talking about communities of practice. So from the literature, there's teaching tech together is a great book by one of the founders of the Carpentries. I really recommend taking a look at. And so to use practice are people bound together by interests and an activity such as knitting or particle physics. And one really key piece here that I think I always wanna highlight is this idea of legitimate peripheral participation. So this is this idea that you can be part of this community, can become part of this community, not just by sprinting in and being like, ah, yes, I too can do exactly what this person up here is doing, I am the same. Because honestly that can be very difficult and demotivating. Instead, these communities tend to work really well when they can help people who are maybe tentatively walking into this space meaningfully participate. So in this example, it's saying that maybe you're not, you know, you're not immediately directing this project but you're starting, you're participating, you're making your first scarf, you're stuffing envelopes, you're proofreading documentation, that's huge. And I think that this is a really helpful reminder for how it feels when we're new to these communities that we really love and care about is that giving people about to meaningfully participate. The other piece is something that I really appreciate from the Arcus Education team, kind of the mission statement, we really focused in on this model of learner instructor. And so from a mission statement, we try to focus on creating opportunities for community members to develop as learner instructors who learn, teach, coach and mentor their colleagues and community. So we understand that pretty much all of us have experience with being the learner and being the instructor. And we kind of rejected the idea that there should be any divides at all. This is another way that we kind of think in about, you know, we have these extant communities, these different communities of practice, but just as we wanna make sure that newcomers can come and not feel like this is cool, but this isn't me, at the same time, we can kind of grow and think about our communities as being spaces where everyone, no matter where you're at, can lead and teach others. Everyone has something to share. So that's an important kind of idea for us. And so of course, the next half of this, I'll talk a bit about what it was like, you know, once COVID had hit. You know, we basically being at this, you know, research institutes part of a hospital, of course, like basically shut down of anything that was an essential work being in person. So a lot of the research institutes went remote. You know, there's a lot of distinction. If you weren't, if you didn't need to be there, you know, you weren't visiting on site. Now this created a really important impact where people were basically, who might have had projects on site or research that was not essential, but very important, were basically sidelined for a while because they couldn't get into their physical resources or experiments. And so there was a real surge of people who were all of a sudden in this kind of precarious position of not really knowing how to spend this time. And so our fearless leader, who you'll hear from next, Joy Payton, had the great idea to raise her hand and be like, well, why don't we, why don't we do some education? Why don't we hold this webinar series? We'll call it lab down skill up. So we'll basically create a rapid, wide-scale data science training program. And let's just start it in, I can't remember if it was one day or two days later. And so we basically agreed to teach three webinars a day. And this is what our spring and summer looked like. It was wild. We taught over 130 sessions. And when I say we taught, I actually mean going back to the learner instructors that we facilitated, you know, this process where 31 of our, ourselves and our colleagues, largely technical folks who may or may not have had backgrounds as educators, were the ones up there leading these sessions. And we also really committed to this regular cadence. You can see about halfway through, we really, we took a break for a couple of weeks and slowed down and we went a little bit more reasonably Monday to Thursday, two times a day. But we also, you know, we noticed that people were indeed coming a lot and coming back because there was this element of familiarity and structure and this personal element that people really connected to. So we were doing stuff. And I think that the critical moment came for us a couple of weeks then when people were coming and having this seemingly positive experience. And also we realized this is a huge opportunity to understand what people are going through now, but in general, what is the state of data learning and data needs at our institution? Like how do we go from this massive activity to learning about what it's like to learn here? So we said, let's slow down, take a couple of weeks and let's ask people, you know, let's conduct some research. So, and this is kind of what one unit looked like. Jake Riley was leading a session here on intermediate data visualization. We created content that we can use later. And these are some of the statistics about, you know, attending and I will just keep going and talk a little bit about evaluation for the remainder here. So the key pieces here, we use mixed methods research. We were able, I'm realizing I want to make sure to have time for everything. So we basically had surveys, we had user interviews. We ended up looking at who was there and we noticed that it wasn't just data analysts or it wasn't just, you know, one role. It was actually this really interesting mix of people who had already identified that they had basically data in their role. They were analysts. So also people who were sitting as managers, as coordinators, as administrators. So we were reaching people who, we didn't necessarily have the PIs in the room. We had a lot of people who were starting their careers at CHOP or figuring out how they fit into things and that became really, really important. So as we looked at this, we learned that people really responded to the welcoming atmosphere here. So this was really something people, more than anything else, they responded to the instructors as people up there, you know. And that was really neat to see that people wanted this sort of experience. They want, there was this emotional element that was really important at the time. We also learned that people really wanted more guidance and sequencing, which was a very key actionable insight. When we think about kind of tactics, we realized that we needed to give more description, more context so people could move through this. So we had this emotional element in terms of being a reliable community space, but we also wanted to make sure to structure the content so it was accessible. So these are things we're learning from surveys. But maybe one of the newer things that we realized was that the social context here was absolutely key because what was happening was that people were not just taking information as end users but trying to figure out within their units, within their part of this institution, kind of what it meant to learn and grow. And so we had people, and it wasn't always the same answer. You know, we had folks from program managers talking about the need to create trainings for groups of people. It wasn't enough just to assume people had the buddy, right? So we couldn't just assume that people already just know a person. We need to be explicit about creating these opportunities. But then we also had folks who were ready to go, perhaps interested in joining these bigger groups like some of the user groups, and they knew that these were positive experiences, but they were just intimidated by that and they felt much more comfortable by having that buddy. So if you have the buddy, it's great, but if you don't have the buddy, what do you do? How do you make this all work? We also found out that a lot of people really needed to convince their supervisors of value. And so I love this idea here, which is that somebody in an interview said that, the boss said that if I had the magic button, that would be great. And so I made the magic button. So in this case, something like dashboard reporting was key to convincing the value here of what's going on. So what I wanna just leave you with here is this idea that when we think about communities of practice, where you think about this actually goes really well with the NSR idea of going from the flat communities and non-hierarchical to actually understanding the hierarchies or I'll even just say the networks across communities. Because when we think about supporting people, we might wanna also imagine the idea of a network. And this kind of sets our future research agenda. This is something we're working towards. Is that we want to think about how it's influence plays out within a network. So this is a network visualization with nodes that are connected here. And so when we think about things like influence across a community, you can think about ties within a community, communities becoming more intermeshed. This is something from Washington University. But we can also think here about, if we wanna grow communities, how are those communities actually structured? You know, it was a network graph, a useful metaphor here. So we think about how to support people, how to help them step up and how to create materials so that for like the nodes within the network, so that we can actually help people get protected time. We can help people who don't know the person reach out to you and especially the folks who might not, you know, share some of the same normative identities, the same access, the same power within an institution. You know, folks who are coming into a research context and they're ready to learn, but they just don't have that access yet. We can help create that by being strategic about what we learn about communities. So this is really a, I would love to see more of this type of work. And I'm a mixed methods person, so I will always love the idea of the network. That's something that just calls out to me. But I think it provides a nice layer of complication because we work in really complex institutions. So how can we kind of think about going from communities here to these great opportunities like webinars and events to really advocating for people's growth and movement within these networks? So I got a little rushed there. I apologize there, but I am really excited to share this because I believe a lot in strategy. This is how I'm growing a lot personally within my work and data is finding ways to celebrate vulnerability, kind of having mental models that work for understanding influence. And then creating these meshes of communities of practice. And then finally, never forgetting about what it's like to be a newcomer and making sure that there are meaningful ways to you participate as newcomers. So those are some of the ideas. I would love to hear any questions. And here's my cat and I reunited in the end, so. So I think we've got time for one brief question. And the top one was, how do you keep people engaged when you're doing these presentations virtually? Oh, it's a wonderful question. I think going back to our instructor kind of toolkit is meaningful participation, not putting people on the spot to have to share their video necessarily. But I think, because we're all kind of fatigued, right? Except me, hi. But I think really having meaningful ways to ask people questions in the chat. I love having people predict. That's one of my favorite tools here. Being like, why might something break the kind of Socratic method? And just making sure that there's meaning that people feel really included in meaningful ways. And you can even be strategic and be like the beginning at the middle of the end. I wanna make sure to have asked enough questions that people can jump into. I love chat for that personally. Yeah. Yeah, this has been really great. Got some good ideas here to think about some of those too. And we're gonna then, I think switch to the next session. Perfect. Sorry, thank you everyone. Yeah.