Episode published: Friday 12/27/2024
Michael: Hi everyone, and welcome to a brand-new season of Every Day is Groundhog Day (Except for the Days When It's Not), the only podcast devoted to the best holiday there is, Groundhog Day. If you're new to the show, welcome. Feel free to go back and listen to some of our Season One episodes. We have shows devoted to Groundhog Day forecasters like Sylvia the Apex Armadillo and Gertie the Groundhog, and topics like Groundhog Lodges. I'm really excited to be back talking about Groundhog Day once again. We have a lot of great guests lined up for our second season, and I can't wait for you to hear all of them.
For today's episode, I spoke to Dr. Eric Green from Duke University. A few years back, Dr. Green reached out to me to see if he could include the Groundhog Day forecasting data that I've been collecting on CountdownToGroundhogDay.com for use with the data package that he was putting together. This data package would be used as part of a fun project he wanted to have his students work on near Groundhog Day. I agreed that he could use the data, and he's now had a few of his classes do this project. I wanted to learn more about what he did with the data and his work at Duke University in general, so I asked him to come on the show. So, without further ado, here's the interview with Dr. Green.
Michael: So, today I'm speaking with Dr. Eric Green, who is a psychologist and professor at Duke University and who, back in 2023, created a data package combining the Groundhog Day predictions that I collect for Countdown to Groundhog Day with historical weather data from the National Oceanic and Atmospheric Administration. He teaches a data science class where he has his students use this data to determine which prognosticators are most accurate. So, welcome, Dr. Green.
Dr. Eric Green: Thank you.
Michael: So, I want to talk about the data package and your class project, but I guess I wanted to ask about your background and your field. I know you said you are an Associate Professor of Global Health and also Director of Undergraduate Studies for Global Health as well. So, what does that entail? What's the focus of global health?
Dr. Eric Green: Global health can really focus anywhere geographically, but it looks more at this issue of equity and inequities in being able to access health care and differences in health outcomes. So, an example would be during COVID, you know, not everybody is equally winding up in the hospital from COVID in the early days; it was largely Hispanic populations compared to white populations in the US, right? And so, that's an inequity in a health outcome. That would be a typical global health focus there, to say, why are some groups experiencing this at higher rates than others based on what you'd expect from them in the population?
Michael: And is that something like their jobs in that case? Like, is there a higher proportion doing the in-person delivery type jobs, things like that? Is that kind of what you're focusing on or trying to figure out?
Dr. Eric Green: Yeah, that's where epidemiologists might note this trend, and then other folks in global health study this and figure out like, what's the reason behind these inequities? And so, global health brings together people from many different disciplines because these challenges are complex. There's social and structural reasons why; you mentioned one of them, it's your job exposures, right? So, some people are more likely to be in riskier jobs than others, and that's related to economics and education and, kind of, background. So, global health brings together folks from the humanities and medicine and statistics, and I happen to be a psychologist but trained on the quantitative side. So, there's lots of homes in global health.
Michael: And how long has this been, like, fields or an area of study? Is it something… I don't know that I remember it from, like, when I was in college. Is it something kind of recent and, you know, last couple of decades?
Dr. Eric Green: Yeah, it's gone through different names and a different focus. In the past, it might have been tropical medicine, just looking at diseases in the tropics, and then international health, which is looking at comparisons between countries. And yeah, when you and I were in college, there were no… You couldn't major in global health, right? Actually, Duke has one of the first global health majors, it was around 2012 or 2013 that we started offering global health as a major. So, as a field of study that our students are looking into, it's still relatively new.
Michael: So, I guess I should ask what your connection to Groundhog Day is before we get into the package.
Dr. Eric Green: Well, when I was 3 and 4, maybe I was up to 5, we lived in Punxsutawney. I grew up for part of my childhood, kind of just at the cusp of my memory, really. But my parents have really good friends in Punxsutawney so growing up, we would go to Punxsutawney several times a year. I never made it there on Groundhog Day but it was always a part of childhood growing up and my mom and my parents are still connected very deeply to that part of the state. I grew up in Harrisburg and originally, from Pittsburgh. So yeah, it's just been something I knew about as a kid, and growing up in Pennsylvania, it's definitely… Everybody knows about Groundhog Day for sure. So, that's my connection to it.
Michael: Okay. Did you like the movie?
Dr. Eric Green: Yeah, I've seen the movie a bunch of times for sure, and talked with people in Punxsutawney about the filming of it and their kind of recollection of that. Definitely a classic and who doesn't love Bill Murray?
Michael: So, do you have any hopes of getting out to see the ceremony or have you kind of given up on that?
Dr. Eric Green: Well, we livestream it in our house now, so I'm trying to get my kids interested. I don't know when this schedule is going to work out. I'm usually teaching that semester, but I would love to get my kids up there. They're still young, almost 10 and 6, but I feel like they would get a kick out of it. But they love watching the livestream.
Michael: Yeah. So, we do that too, or we have in some years. The last couple of years, I've been trying to get out to some of the other ceremonies in the area so sometimes we're on the road. I've brought my kids to some of them. So, sometimes they get to skip school. But this year is Sunday, right? I believe it's Sunday that Groundhog Day falls on, but I still don't think I'm going to try that. I don't know that I can take that cold. But I feel like the time for that would have been like when I was in college, maybe, because I went to college in Pennsylvania.
Dr. Eric Green: I went to Bucknell in Pennsylvania, and it was definitely a missed the opportunity to get out there. So yeah, I agree.
Michael: Let's talk about this data package that you put together. When did you come up with the idea? I know this is part of a class and I guess I should ask, is this something you're going to be teaching this year as well? Or do you have that data science class coming up in the spring?
Dr. Eric Green: So, this course started as a reaction to COVID. Back in the spring of 2020, our students went on spring break, and we told them not to come back. So, all of a sudden, courses interrupted, students, I mean, everyone's at home and I decided to offer a virtual, weekly get-together, ‘workshop' is probably too strong a word. But we had 60 to 70 people; this is people from Duke connected to our global health institute, faculty. People were looking for something to do, this is when you were excited to jump on Zoom, right? And it's maybe not like how it is today. But really, I think people were looking for connections and I offered this class as a way to learn the software package R for data science. And we were looking at COVID data. So, this was on everyone's mind, and it started as a remote workshop in the spring, turned it into a summer course that summer, and then after that, it became a regular semester course.
And then, you know, move the focus away from COVID, I was just kind of burnt out on looking at COVID data myself and on my students weren't going to be interested in that long term, so I made it a more of a general data science course, but with a focus on global health. You know, I try to keep the examples really tight to health, but I was looking for something fun as a way to teach data wrangling and also how to get different sources of data from around the internet. Groundhog Day was coming up and certainly, I have this connection to it. I was looking to say, "I wonder if there's any Groundhog packages out there." And I was looking online for Groundhog data, where can I get information on this? And I stumbled on your website, and I was so surprised to find that there are so many cities and towns out there doing a Groundhog Day festival and you had this historical data that in a weekend I started thinking, "Okay, this has got to be the project we do." And then I reached out to you, and you were kind enough to let me scrape your website for the data and that's really how I got started.
Michael: Did you scrape it, like, with Python or something like that?
Dr. Eric Green: Everything was within R. I used different R web scraping packages. You know, your website is really set up in a structured way so it didn't require a whole lot to be able to go in and get that. And yeah, so it was pretty easy to put together for the package.
Michael: Yeah. So, I'm not familiar with R and when you reached out to me, I was like, I should probably look into it because I would like to play with the package that you put together. So, that's still something on my list.
So, could you talk about what the actual project was? What did you ask your students to do?
Dr. Eric Green: So, every week for the first few weeks of the semester, I have them participate in this kind of online community called Tidy Tuesday. And folks in the R community, it's one of the best software communities I think you can be a part of, they put out these weekly challenges. They share a data set and essentially, what the challenge is just to do something interesting with the data set and make some visualizations (it's really tuned toward tables and visualizations), share them online, and ideally, share the code you're using to work with these open-source data sets.
In my course, I'm teaching students how to be better, get better at visualizing data. And I think it's really great to have them take a new data set, think about what they'd like to do. In the beginning, their skills are small, so they're making very basic plots, they get more complex as the semester goes on, and they're seeing what other people are doing as sources of inspiration. I knew that once I realized all this groundhog data were there, that this could be something that we took on as a project one week, but then it ended up… It stuck around for a while because there's a lot of data there with the weather data and prediction data. And so, I had my students take this on for a couple of weeks as a project to look at data wrangling and data visualization.
Michael: You had them compare… One of the things I think was to try and figure out which prognosticators were most accurate. Do you remember the outcomes? Did you yourself run these analyses? Do you have the results somewhere or was it mainly the students doing this?
Dr. Eric Green: It was mainly the students. I do remember running a few myself, like, showing that if you look at years with at least five predictions, the percentage of early spring yearly predictions tends to be on a decline from 1980 up to now. So, across the whole data set, it's less and less likely that prognosticators are predicting an early spring. That's one thing, as a share of the total predictions.
Michael: See, that's interesting because I feel like the last couple of years, it seemed like a lot more were coming out with early spring. I was wondering if that was just like, you know, some of these newer ones just want to give a more positive prediction.
Dr. Eric Green: My students have, you know, they've made fun visualizations that, you know, use images of groundhogs in the charts themselves and, you know, look at by prognosticator type. One of the things I did was took all of your prognosticators and I think we labeled them by "animate" and "inanimate" because I was really surprised to learn that a lot of cities and towns, some of them were using live animals, others were using stuffed animals or statues or humans dressed up as animals. So, we were able to code the prognosticators by these different categories, inanimate or animate, and then further code them by their classification in the animal kingdom, essentially.
Michael: Right. Yeah, yeah.
Dr. Eric Green: We had a lot of fun with it.
Michael: Yeah, there's a lot of there's a lot of live groundhogs, but then you also have a lot of stuffed groundhogs and then you have different animals. There's definitely, like, animatronic groundhogs, there's puppets. I know there's at least one… I feel like there might even be a couple statues at this point that people use. So yeah, you could definitely get into a number of different ways to classify live groundhogs versus dead groundhogs versus, you know…
Dr. Eric Green: I will say the prediction accuracy among inanimate prognosticators has been on the rise since 2005, so they're catching up and even surpassing the ability of creatures.
Michael: Oh, wow.
Dr. Eric Green: As a learning exercise, you know, there's not a straight-line connection to global health. Certainly, you know, as we look more at climate change, there is something here, but it's been a fun exercise because it brings together prediction data, which means you can look at accuracy, questions of accuracy. To do that, you have to create some of these variables by bringing in the weather data, right? So, to see if a prediction ultimately ends up being correct and you have to make a definition of, what does correct mean? So, for the students, it seems like a really silly kind of fun exercise, but from a data science perspective, it brings a lot together, different data sets of different types and a lot of data wrangling. So, the students have a lot to work with here.
Michael: I think you said in one of our emails, am I remembering correctly that your mother brought in cookies one year, like, groundhog cookies for the class?
Dr. Eric Green: Yeah, she lives right down the street from us. She offered the first year to make some groundhog cookies, and then it expanded to all of the classes I teach, not just the data science class. So, I think she's making multiple dozens of these cookies, and she packages them great and brings them in. You know, our Master of Science and Global Health program, it's where a lot of our students come from for this data science class, they are 50 or 60% international students. The premise of this assignment is just pretty crazy to them, right? You know, I say, "Okay, well, we're going to look at a data set where a groundhog, but also other animals and mascots are predicting the weather." And so, they're looking at me like, "What do you… What did I sign up for?" And, you know, they walk into class, if it happens to overlap on Groundhog Day, they walk into class to a recording of the livestream, right? It's just a cultural experience for them, and then I pass out my mom's cookies. Every culture has some weird things, right, "Let me tell you about this one." So, they get into it too.
Michael: Yeah, that's interesting, because it is mainly a North American holiday and kind of centered around the Pennsylvania area, I mean, I guess outside of the country and maybe in Canada, unless they've seen the movie, it's probably, "What is this?"
Dr. Eric Green: Yeah, this is totally foreign to most of our international students. I think it makes it part of the fun.
Michael: Yeah, that's great. Hopefully, they'll spread the word, let their friends and family know about it. I do look for ceremonies that are outside the country, and I feel like I've seen ones that… You know, maybe Russia has them sometimes, somewhere in Russia, I feel like I've seen it. But it's hard to find it outside of.
Dr. Eric Green: Well, I wanted to ask you: How do you find out about these new ceremonies?
Michael: So, a few ways. I would say it was probably around 2018, 2019, I had mainly been focused on Punxsutawney Phil, but I knew Staten Island Chuck, and I knew about some of the some of the ones around here; we have Essex Ed, used to be at the at the Turtleback Zoo, which is here in New Jersey, and now it's Essex Edwina. But I knew that I knew there were other ones, and I just decided I'm going to put a blog post together for my website, just listing out the different prognosticators. And so, I would start searching and I would find the occasional list and it would be like, you know, "Here are the 10 other Groundhog Day forecasters," and it would almost always be, like, nine of the same and then there would be one different one. And as I slowly went through this, I was like, you know, there's a lot more than I thought there were going to be.
So, I got a list off of Wikipedia or something like that of zoos and farms or something like that, and I just started reaching out to as many of them as possible and being like, "Do you have a Groundhog? Do you have a forecaster?" And as many of them as I could find, and I got a lot of that from that. And then I also have a Google Alert for Groundhog Day, which most of the time doesn't give me anything useful because it's "Somebody lost a basketball game, it's Groundhog Day again." Sometimes it'll turn up some forecaster that I don't know. People do write in to me occasionally and tell me too. So, it's a whole lot of different ways. I haven't figured out any great way. Like, I don't know if AI can help me, if I could have something go out and just like scrape everything and try and find it. But it's tough because some of them are really big, these ceremonies, you know, and then sometimes it's just some local thing that somebody is doing in the town and maybe it's not getting coverage, maybe it's not getting in the newspapers or maybe it's just like, you know, a post on social media.
Dr. Eric Green: Well, I mean, you've surfaced so many of them. It's got to be a really time-consuming process to then go in and document all of their predictions.
Michael: Yeah. So, Groundhog Day has gotten really busy for me. I try to go to some ceremony at least in the morning, one or two, and then the rest of the day is going back to all the sites that I've collected, they usually have some social media things. So, it does it does take time. It's fun, I enjoy doing it. I enjoy, you know, seeing all the results as I slowly update my database. But, you know, sometimes my kids help too.
Dr. Eric Green: [chuckles] Well, you have some of my favorites that I've learned about, you know, from my hometown in Harrisburg, Sheena, RIP, sadly departed, the dog dressed as a bear pretending to be a groundhog. Got to be a favorite.
Michael: Right. Yeah, that was one of the early ones I found, and that probably helped inspire me to be like, "I've really got to I really got to try and document all these things," because that was an interesting one. Yeah, unfortunately, she passed away a few years ago.
Dr. Eric Green: Too soon. Too soon, for sure.
Michael: And I think her owner, her owner moved to New Orleans, and I know she did like she had an alternate at least one year, but I don't think she did it last year. So, I don't know if that tradition is done or not.
Dr. Eric Green: I've also looked at the North Carolina prognosticators. And again, you know, too many of our departed prognosticators, Sir Walter Wally was in Raleigh, and there's not too many others in North Carolina. So, it's not really moved this far south as a tradition, a common tradition.
Michael: Sylvia is from North Carolina, right? The Apex Armadillo?
Dr. Eric Green: Yes, the Apex, right. Yes, yes, in Apex, which is not too far from us in Durham.
Michael: So, maybe if you can't get to the to the Punxsutawney, maybe you could get out to that one. I did interview the mayor who came up with that tradition last year for the podcast, Sylvia, the Apex Armadillo, I think it's been, like, four or five years now. Yeah, that's an interesting one, too.
Dr. Eric Green: I might have, as we're talking, just created your newest groundhog. I don't know, let me know. I went into ChatGPT and I said, "Imagine that you were a groundhog that makes weather predictions and it's February 2nd, 2025. What would you predict for the spring: early spring, or long winter?" And ChatGPT tells me, you know, "As your trusty groundhog, I come out from my cozy borough and today I see no shadow." So, maybe this is your first February 2nd prediction and first AI prediction.
Michael: Can you make it give itself a name?
Dr. Eric Green: Oh, yeah. Okay, so… "Give yourself a name as the newest and maybe first AI prognosticator." Let's see. "I shall be known as Byte Burrower. My digital sensors may not rely on shadows, but I'll still bring accurate predictions for the changing seasons. When February 2nd rolls around, the world can trust Byte Burrower for timely weather insights. No burrows or hibernation needed." So, there you go.
Michael: That's pretty funny, yeah.
Dr. Eric Green: So, noted here, no shadow. [laughs]
Michael: No shadow. All right. We'll have to see if that holds true. So, let's see. Is there anything else you want to mention about that project or your data package that I haven't asked about?
Dr. Eric Green: Well, I would love for anybody who's listening, maybe works with R, and knows the NOAA data. The package that I rely on is in the middle of undergoing some changes because NOAA changed its API, right? And so, the API is the way that our packages connect to their data. They've changed that API. So, the R NOAA package I relied on won't pull the most recent data. And I think some people are working on this. But that's one area where, in order to get it back up and running for these most latest years, being able to connect to the NOAA data from R is such a big thing. If any of your listeners know about that.
Michael: Okay, yeah. We only have a couple of months, that's worrisome. Yeah, I was going to ask about that, how you got the data? So, it's through an API or it was through an API, or a package, that uses that API from NOAA?
Dr. Eric Green: Yeah, exactly. There was a great R package called R NOAA, and it made it really easy to go in and grab NOAA's data on all of the weather stations and all of their historical daily max temperature data. So, you know, connecting that with the lat and long of the weather station and the lat and long of the towns where the groundhogs are from, it makes it all possible to put these relational data sets together. But the connecting to the R NOAA package is going to be key.
Michael: I see that you co-founded a company called Nivi. Could you talk a little bit about that? Is that something that's still ongoing or is that, like, a side thing?
Dr. Eric Green: Yeah. So, when I finished grad school, I did a postdoc in New York at the Program for Survivors of Torture at Bellevue Hospital in NYU and then moved up the street to a health research group called the Population Council and there, led a team that got a grant from the Gates Foundation and USAID and a few other organizations to develop a screening and referral tool for pregnant women and new mothers in Kenya to use. I approached that very much like a researcher. I was a researcher, which means that when the money ran out, the idea was done because I didn't have a business plan and any real ideas.
I met my eventual co-founders of Nivi and decided, "Hey, we think there's an idea here. We think that screening and referral, being able to do it on the phone, helping people to connect to local health resources and get good information, high-quality health information is really a great idea." And so, we found funding to really kick the tires on this idea. Merck for Mothers came along and gave us some startup seed funding. At that time, I had moved to Duke, and we decided, with Duke's help, to spin it out into its own company called Nivi. That was in 2015, 2016.
It's still going. Nivi works in Kenya, in Nigeria, and India as the main markets, kind of, further expanding as a team of 15-20 folks in all of these markets who are working. And really, I really had not been part of something that lasted as long in public health as this company. So, that was surprising to me. If you'd asked me in 2012, you know, "What's the best way to have an impact in public health?" I would have said, "Well, work on this grant or publish these papers." And, you know, those are great contributions, but really working with Nivi was the longest-lasting thing, initiative, that I had been a part of. It was kind of free of that grant cycle where you kind of wrap up things in two or three years. I worked with some great co-founders and colleagues that really pushed it forward and I learned a ton. You know, all I knew about business going into it was from Mark Cuban on Shark Tank. So, this was a real learning experience for me.
Michael: Cool. That's pretty impressive.
I think that covers pretty much what I wanted to cover. I'm glad that you've put this together and I'm glad that your students seem to enjoy it and that you're definitely spreading the word about Groundhog Day.
Dr. Eric Green: Well, thanks again for making your data available. I know it's a lot of work and I really teach my students about the importance of open data sets, open software, and the things that are possible when people share their work. You know, you don't necessarily think of all these things when you first made the data set public, right? And that's how it works. So, the students learned a great lesson from this, I think. So, thank you.
Michael: Oh, thank you. I'm glad you find it useful. I will link to the project in the show notes if anybody else wants to go and experiment with it. And I hope the NOAA data gets updated, the API or the package before the next Groundhog Day.
Dr. Eric Green: Let's see how Byte Burrower does come February. You know, the first AI prognosticator, right? I think we made some history today.
Michael: We did make some history, yeah. We'll have to see how accurate that is. I'll keep it in mind and see if it's correct.
Dr. Eric Green: Very good.
Michael: Thanks so much for talking to me.
Dr. Eric Green: It was great to talk to you.
Michael: So, that interview took place at the end of the summer. As I was editing the discussion, I was curious if there was a way for me to converse with Byte Burrower, the AI prognosticator that was born during that discussion, or at the very least, if Dr. Green still had access to it. I sent him a note on Instagram, and a few days later, he replied that he had developed an app which would allow anyone to talk to Byte Burrower. I wanted to learn more. So, we got on another call to discuss this latest development.
Michael: Okay, so I am here once again with Dr. Eric Green from Duke University. We spoke a few months ago. And during that discussion, you had the idea to ask ChatGPT if it had a Groundhog Day prediction for the upcoming holiday. And it gave you one, correct?
Dr. Eric Green: It did. The prompt was that, "You are an AI Groundhog, and I need you to make a prediction." And it did. And then I asked it to name itself, and it named itself Byte Burrower.
Michael: And then I guess a few weeks ago, you had the idea to make this into its own app that you could get to directly from a website, right?
Dr. Eric Green: Yeah. When you messaged me on Instagram and said, "Is it available to anybody?" I thought, well, no, but I need a good use case for another course I'm teaching in the fall. It's going to be a course, kind of, AI tools and AI usage in global health. For me, I have a background in statistical programming, but I'm not a web developer. So, it was a great use case to see if I could use some AI tools like ChatGPT and Claude to take this idea and basically project manage it. I would be the idea guy, the project manager, and let Claude make the website. And it really worked! I was amazed. So, I just kept prompting it, telling it I needed to connect to a few weather APIs, and I needed to connect to ChatGPT to be able to get the predictions, and I asked it to walk me through the process of setting up a web app to be able to do this, and it did. Maybe it was really keen to see an AI friend come alive on the web, so it helped me out.
Michael: So, you mentioned Claude. I'm not familiar with that tool. Could you discuss it a little bit more?
Dr. Eric Green: It's a competing LLM and it works… If you've used ChatGPT, a lot of folks have at this point, it's very similar, you give it a prompt. And I really found that I liked it better for this task because it gave me feedback and kind of really bite-sized chunks. So, "The first thing we're going to do is we need to set up the front end of your website, so we need a place for people to say what city and state they're from." And so, it walked me through that and then it said, "Okay, now that people can submit a city and state, we needed to do something when they click submit. So, let's connect to a weather API." So, then it told me how I would do that. So, it kind of walked me through step by step until I had a fully functioning website. I even went back, I think I switched back to ChatGPT here and said, "You named yourself Byte Burrower. I'd like a logo." So, then it came up with an AI groundhog.
Michael: On the Byte Burrower website, you mentioned that you're incorporating the historical Groundhog Day prediction data from CountdownToGroundhogDay.com. How do you get the data into the app?
Dr. Eric Green: Yeah. What it's doing is it's grabbing the data on the prognosticators once, you know, since the data are only updated once a year, essentially, it only has to do it once, it's not a daily thing it needs to do. But it can grab the prognosticator data once, it can grab their prediction data once. And what I'd like to do is, you know, somebody says that they're from Harrisburg, Pennsylvania, or Durham, North Carolina, that it would not only give you a prediction for February 2nd, what the weather, you know, for an early spring or long winter for your location but I'd like it to also tell you who's your closest active prognosticator, right? There are so many in your database. So, just looking at where a user is, who's the closest prognosticator, and then maybe giving them some information about, historically, what have they predicted and how often have they been right about it.
Michael: So, do you have any of your students working on this? You said this is for a use case.
Dr. Eric Green: Yeah, for a different class. So, I teach a class in the spring, typically, that is a data science class. I am fortunate to be on sabbatical this spring, so I won't teach that class this spring. But in the fall, for incoming Duke students, I'll be offering a class that is about AI and global health. And that class is going to be very project-oriented. I'm going to take students that, most of them will not have a background in programming of any kind, software engineering, or statistical programming, and we'll introduce them to different the ways that AI is used in global health, is being used, and some tools for doing just the type of thing that I did of having an idea and knowing that there's data out there and thinking about how to make your idea come to life.
Michael: Awesome. Is there something you're working on in your sabbatical?
Dr. Eric Green: Yeah. You know, one objective of the sabbatical is, for myself, taking a pause and figuring out how to incorporate AI into my classes. So, this really does just fit within that broader theme. I think a lot of professors, all of a sudden, you know, you're in a new environment where you ask students to do any kind of a writing assignment, and you've got to just assume that it's partially AI-assisted now. And so, you really have to rethink what kind of assignments do you create for students to make sure you understand their knowledge is not some AI's knowledge of a topic. But I don't just want to focus on limiting students and trying to figure out who's doing real work and who's doing fake work. I'm much more excited about how to teach students to use AI to be much more productive and creative and significant in the work that they're doing.
Michael: Great. So, that means that you're not going to be doing the Groundhog Day project this year, I guess.
Dr. Eric Green: My class won't be doing it. But of course, I'll be following along and trying to make sure that the data get updated when you post your heroic efforts to gather all of the predictions.
Michael: Yeah, I've actually been spending some time the last few months trying to update my historical data. I don't know if you've got those yet. There's more and more, I think I have something like 200 overall prognosticators from, you know, Punxsutawney Phil on.
Dr. Eric Green: Do you just have a lot of Google Alerts set up? How do you how do you go about it?
Michael: So, I do have Google Alerts. A lot of this research has been done by going to one of those digital newspaper archives and I'm just looking back, I'm trying to find mentions of other groundhogs when Punxsutawney Phil is mentioned or some of the other prognosticators. If I look up, like, "Dunkirk Dave," sometimes it'll say, "Oh, here's some other ones that are currently doing it," and these are ones that I've not necessarily heard of because they were 30, 40 years ago. So, I discovered one that's, like, there was Mr. Prozac or Zac, who was like a llama who predicted the weather for maybe, like, six or seven years back in the early 2000s. I'm discovering other prognosticators and other types of animals that have done this that might not necessarily have a counterpart currently. So, it's slowly growing. It's a bunch of work.
Dr. Eric Green: I'm so glad you do it because as a teacher, it's just turned into a fun way to teach current challenges. You know, there are there are other real and mock data sets out there, but they're often not very interesting. And sometimes it's tough to jump right into the use case of, you know, pharmaceutical data or global health data because it's hard to understand the data and the process at the same time. But this groundhog example is nice because it involves historical data, longitudinal predictions that individual prognosticators have made, it brings in the geography of it. You can connect to different APIs because it's related to weather, and it's related to predictions. It actually packs together a lot of the things that you might want to teach in a data science class in a fun example. For me, to teach it in the spring and Groundhog Day happens in the spring is a nice coincidence too.
Michael: Awesome. Is there anything else you want to mention?
Dr. Eric Green: No. Again, thanks for working on the project. It's super fun.
Michael: I'm glad you find it useful. I don't know how many people do in the world, but at least we got one.
Dr. Eric Green: You got one. I'm sure you got two somewhere out there, but you got one. [chuckles]
Michael: All right. Thanks again for talking to me today.
Dr. Eric Green: Of course, of course. Have a good weekend coming up.
Michael: And that's today's show. Thanks again to Dr. Eric Green for speaking with me for today's episode. If you want to try out the data package that he put together, combining the Groundhog Day forecast data from CountdownToGroundhogDay.com and NOAA's historical weather data, there's a link in the show notes. Also, you can go to GroundhogDay.app to visit Byte Burrower anytime you wish.
Music for this show was written by the fantastic Breakmaster Cylinder. Show artwork is by Tom Mike Hill. Transcripts are provided by Aveline Malek at TheWordary.com. If you want to learn more about Groundhog Day, visit CountdownToGroundhogDay.com. Any feedback or voice messages can be sent to podcast@countdowntogroundhogday.com. Thanks for listening! Talk to you next time.
--------
Transcribed by Aveline Malek at TheWordary.com