Deep Dive: A Shared Approach to Data Governance
This episode is brought to you by Veracross and Toddle.
The Downtown School is a small school that takes a unique approach to school management – all faculty and staff also serve an administrative role. In addition to a strong sense of ownership, the approach lends to an immersion in school data by everyone, leading to better strategy and positive outcomes.
Data security is especially critical, including when using AI to assist with data analysis (though there are ways to do so responsibly, e.g., via API and special arrangements with the AI providers).
The model is scalable (e.g., a department could adopt this approach as opposed to the entire school) as well as flexible.
Also discussed was the work of the Center for Institutional Research at Independent Schools (CIRIS) and new AI-related initiatives they have coming.
Resources
- Harpertech.io
- The Downtown School
- CIRIS
- Data-Informed Decision Making: A Guide to Institutional Research in Independent Schools
- Canvas by Instructure
- ChatGPT
- Claude AI
- FERPA
- COPPA
- NAIS events
- ATLIS AI Resource Hub
- ATLIS content - AI
- ATLIS content - Independent School Governance and Leadership
- Visit Seattle
Transcript
Narrator 0:02
Welcome to Talking Technology with ATLIS, the show that plugs you in to the important topics and trends for technology leaders, all through a unique independent school lens. We'll hear stories from technology directors and other special guests from the independent school community and provide you with focused learning and deep dive topics. And now please welcome your host, Christina Lewellen.
Bill Stites 0:25
Hello everyone and welcome back to Talking Technology with ATLIS. I'm Christina Lewellen, the executive director of the Association of Technology Leaders in Independent Schools.
Hiram Cuevas 0:34
And I'm Hiram Cuevas, the director of information systems and academic technology at St. Christopher's School in Richmond, Virginia.
Bill Stites 0:40
I am Bill Stites, the director of technology at Montclair Kimberley Academy in Montclair, New Jersey.
Christina Lewellen 0:46
Hi guys. How's everything going? Welcome back. It's always nice to do these recordings with you for our little podcast coffee date that we have with each other. How are you guys doing today?
Bill Stites 0:55
I am doing quite well.
Christina, I know you don't often like to kind of put a recorded date or time around when we do these, because we never know when they're gonna launch. But I need to point it out because our listeners don't get to see us the way we do. Hiram is dressed up like Santa Claus for this, which is incredibly distracting, and I'm just going to apologize to the listeners right now, because I'm gonna have a hard time focusing even though we've got a great guest with us today. Hiram is just going to be causing such a huge distraction. I don't know what I'm gonna do with myself.
Christina Lewellen 1:24
Yeah, and it's not even like it's a full-on Santa, with a hat and a nice white beard. It's literally Hiram in a Santa shirt. So you look like a mall Santa. I mean, you're not even a good Santa. You're like a knockoff Santa.
Hiram Cuevas 1:38
Wow, wow. I “live on a throne of lies”.
Christina Lewellen 1:44
“You smell like beef and cheese.” We probably have a copyright situation going on.
You know what y’all, today we're welcoming our guest, Hudson Harper. He's gonna be joining us in a few minutes, and I'm really excited. We cover topics, we cover people on this podcast. But one of the things that I have to start out with before we welcome him to the show is that Hudson is from the Seattle area. I have a friend who told me years ago that everybody has a soul city. And I'm an East Coast girl – I grew up in New York; I live outside of DC. Why in the world would Seattle be my soul city, but it is every time I'm there.
Like my heart rate just goes a little slower; my blood pressure's lower, right? I don't know if it's that, with my gluten allergy, I can eat cinnamon buns or something down at the Pike? They have gluten free cinnamon buns that I absolutely love. But that city just speaks to me. So before we kick off the official reason we are gathered today, I want to know – Hiram and Bill: what's your soul city? Where do you love to visit?
Hiram Cuevas 2:49
I tell you, you mentioned Seattle, and I would say Seattle's my Soul City as well.
Christina Lewellen 2:54
We're besties.
Hiram Cuevas 2:55
Yes, we're buds mainly because my wife and I – we honeymooned out west, and we launched our journey through Seattle. And we love Seattle, absolutely love Seattle. Went up the Washington coast and did the Dungeness crab, the Pike Street Marketplace, the Space Needle – the works. And every time I go back to Seattle, I realize how much I love that place.
Christina Lewellen 3:17
Oh man, Hiram. Okay, so you're my favorite. Bill? Unless you say Seattle. Go ahead.
Bill Stites 3:22
I'm saying Seattle. No, no, no, and not just to jump on the bandwagon.
So those of you that know me know that I've seen all the baseball stadiums in the United States. And our number one trip that we still talk about to this day was when we flew into Seattle, we spent like three or four days in the city itself. We went to a game. We saw the Sleepless in Seattle house, we did the duck boats, we went down along the water, got together with a good friend Peter Baron out there. You know, we did all the crabs, we went to, you know, the fairs – everything. Starbucks, we did every piece of it. We based there, and then we went from there. And we did this trip out to Cannon Beach where they filmed Goonies. And it was just…that part of the country, when we visited, we have said time and time again, if there's ever a place that we would go back to in a heartbeat without thinking about it, it was Seattle, and between the EMP center and just we went to the top of the Space Needle, you know, you go around, people leave notes on the windows – you can do all this stuff.
It was just still to this day, hands down one of the favorite places I've been to and really cool ballpark. They've got this really cool chandelier made of bats, great food. There was nothing about that trip that we didn't like.
Christina Lewellen 4:38
You guys. I feel like we have finally figured out why we have chemistry on the show. Everybody's writing to us and saying “you guys have great chemistry,” “it seems like you're having a lot of fun.” It has nothing to do with, like, astrological signs or birth order, or what you taught in a former life. I really think we figured it out. It is Seattle, our soul cities. This is cool.
Bill Stites 5:01
Indeed.
Hiram Cuevas 5:02
And don't forget the underground tour of the city. That's an amazing tour. If you haven't done that.
Christina Lewellen 5:06
It is a perfect day, I think, to welcome Hudson Harper to the show. Hudson, I'm not sure if you heard all that intro, but apparently we're all moving to Seattle. So, welcome to the podcast. And also we love your city. How are you today?
Hudson Harper 5:20
Thanks for having me. I'm doing great today. And I have to say, I've only been in Seattle for five years, but I think you're absolutely right. It's a great city. I spent the eight years before moving here in New England. You know, Seattle's got very much like Boston vibes, except a lot more chill. Yeah. And a lot less honking.
Christina Lewellen 5:38
And by the way, this podcast is brought to you today by the convention bureau in Seattle, I think. I think we better get some free trips out of this.
So let's get into why we're gathered today. Hudson, I know that you have a great relationship with my co-hosts, and I'm going to turn the mic over here in just a second to Bill, to kick off how you guys know each other. But before we go there, why don't I give you a moment to introduce yourself for real, not just being from Seattle or living in Seattle at the moment, tell our audience who you are; a little bit about your journey.
Hudson Harper 6:09
I am currently the director of technology and institutional research at The Downtown School. This is my fifth year here. And fun fact, it's only the sixth year the school has been open, so, I like to sometimes lump myself in with the founders of the school even though it's like, not really.
I've had sort of a roundabout journey through independent schools, I think pretty much like everyone else, you've had on the podcast so far, you know, it sounds like nobody started off in tech.
In fact, when I was in high school, I was that super annoying kid who was like, “You know what I'm going to do with the rest of my life? I'm going to do pure math that no one understands and talk about with no one.” And you know, I had these lifelong plans, like, “I'm gonna go on to be a tenured professor and live at the same university for like, 50 years.” And obviously, that didn't pan out, at least not yet.
It’s during grad school, where things start to, you know, you realize things aren't gonna go exactly the way that you planned. I was spending about half of my time teaching; half of my time doing research. And for a number of personal reasons, the research wasn't going as smoothly as I would hope. You know, I feel like that's actually the typical experience. But the teaching side was what I started to really love.
And so, short of just extending my PhD program for another, like, three years, I decided – let me explore teaching. And that's when I started looking at independent school jobs.
And so I was very fortunate, I found a great place to work at Loomis Chafee, I was there for a few years, and deciding that boarding schools weren't necessarily the right fit for me, I kind of had this itch in me where I decided, you know, I spent all this time working on a PhD, I want to see if it's possible for me to like, go out and work in industry.
So I started looking around, I got fairly far into an application to work for a three letter agency in DC. I don't know if I can legally say which one it was. But you know, when you're applying for these government jobs, they say, “Don't put your life on hold.”
And so I started looking at other school jobs, and I came across The Downtown School. They were looking for someone who could teach math, but also take on an administrative role. By the way, this was like in June, the year before they needed this person. In hindsight, that was like one of the craziest things I think I've ever seen a school do. But it was a perfect fit. Because I instantly saw, like, The Downtown School, they were really trying to do some innovative, some really cool stuff. And that it would give me an opportunity to explore a lot of the skills and passions I had that went beyond just teaching.
Christina Lewellen 8:38
Yeah. Can you tell us a little bit about the school? What was it about the environment that was kind of innovative?
Hudson Harper 8:43
Yeah, absolutely. I mean, first and foremost, what makes us different is we're fairly small. There are, at most, 160 students in any given year, and only 15 faculty and staff, and every faculty member is also an administrator. I honestly can't think of any other school where that's the model. Everyone has some sort of administrative say or role. Like for me, it's teaching math, but also director of technology. And having been in that system for a few years, whenever I talk to people like Bill or Hiram, I don't see that same kind of dynamics or experiences anywhere else.
Christina Lewellen 9:18
Yeah, it's really unique. I'm sure that there are some challenges with it. But what do you think are the positives of it? If every faculty member has a skin in the administrative game? Is that a positive? Is that a benefit for your students?
Hudson Harper 9:34
Oh, my gosh, I think it's a huge positive. I mean, before we even get to know the students in our classrooms, we get to know them through the admission cycle. So every faculty is also on our admissions committee. And so in many ways, the teachers are the ones who are building are community, and when they start on that day one, it's kind of like, “hey, we already kind of know who you are because we spent so much time talking with you and interviews, open houses, recourse we read your applications.” And I think there's very few other places where the teachers can actually say, “Oh, yeah, we take total ownership of what our school community looks like.” And I think also, once they're here, the benefit is people will sometimes complain about the policies that the administration put together don't always match what the teachers would like in the classroom, and there's no one to blame but ourselves, if there's a misalignment in policy.
Christina Lewellen 10:24
And if I could sneak in one last question at this point – what are the students like? Who's your ideal student who comes to The Downtown School?
Hudson Harper 10:32
Oh, my gosh, the students here are incredibly creative. They're entrepreneurial. They don't mind being at a place that hasn't been around for longer than five years. We'll get into some of the stuff about AI and data warehousing today; I honestly have some students where I feel like I'm trying to keep up with them, because they're so inquisitive, and they're so independent in their learning. They really are just like a special group of kids.
Bill Stites 10:56
That sounds awesome. And I know when you say that you've got kids that you're trying to keep up with, having seen what it is that you're doing, those have got to be very special kids. Because I remember just some of the show n’ tells that you've done that I've been part of and that I've seen, your work is pretty impressive.
And to kind of lean in on that, I mean, you and I had the opportunity to actually work together, really starting probably around this time, maybe a little bit later in the year, last year, when we both applied to be part of the summer CIRIS co-work group, to work on the eventual guide for institutional research that we just published. We all had Eric Heilmann on here a while back talking about that, specifically. But you and I were both one of the cohort members. I think there were a total of eight of us, correct? Yeah, I think that's right. And you and I really had the ability to focus on areas that are at least dear to my heart, in terms of, as nerdy as that sounds, the areas of like data governance, leading into data warehousing, as well as strategies for management of all of that, you know.
I know what my takeaways from that were; I'd be really curious as to what your takeaways were from not only the work that you and I did, but the work that we did, as a cohort, and what you found from all that.
Hudson Harper 12:13
My first takeaway was, one, it's great to talk with you, Bill, pretty much any of our meetings, I was always looking forward to. And along those lines, just like the number of amazing people who are working in this area. I never really thought of the work that I was doing necessarily as purely institutional research. You know, as we were starting up our school, we really wanted to make sure that we were doing things the right way. And in knowing if our new policies if our new programs were being effective, we were just immersed in data and trying to incorporate that into our decision making. So I never even thought about it being as institutional research. And, you know, it wasn't until literally a couple of months before I applied to do this year summer fellowship that I even heard about Eric and CIRIS. And so, like, just getting to know other people who are doing the same type of work as I was, was just a huge takeaway for me.
And also seeing where different schools are at, you know. I think there's a huge spectrum of where schools are in this journey, even if they've started, there's still just a huge variety in what schools have been able to do so far.
Bill Stites 13:18
So one thing I think is interesting that we talked about, you know, with regard to the serious work, but it's interesting given your school, and I'm curious as to how this works, given what you said that every faculty member is also an administrator.
So one of the things that we've talked about is the idea of having like a data committee – people that work within the school to kind of oversee all of those aspects of that. Now, I'm gonna ask you to expound on what those different roles and those different aspects are. But how, given your school, I mean, I can think about the way it is at my school, at Hiram’s school in terms of just our overall size, it's much larger. But in a school like yours, how does that look at The Downtown School, in terms of topics or discussions around data?
Hudson Harper 14:04
Yeah, I mean, in fact, one of our earlier problems, organizationally, was we did literally everything by committee, we were sort of like a de facto data governance committee. And we've tried to pare it down and make that more structured over the last several years.
But I think for me, like one of the big benefits of having the entire administration as your de facto data governance committee, is that it's really easy to hear about what data is important, or what questions do we need to answer.
I think that's something that a lot of schools really struggle with is identifying what are the questions that you need to explore; what are your institutional research projects going to be? I've always had the benefit of, hey, I get to listen in on these conversations. I get to work on this initiative. And it just sort of naturally flows together into, “Okay, this is what we should be working on.”
Hiram Cuevas 14:55
So Hudson, I'm so glad that you're here with us today. I know we've had a couple of exchanges, and we've never really fully been able to catch up in the capacity that I had hoped. And then when I heard that you were working with Bill, I was like, “Ah, this is going to be a great show because anything Bill touches usually turns to gold.”
Now, I do want to touch base on what you just mentioned earlier, which was that continuum, because I know Bill and I are at big schools, you know, 1000 plus students. And I do find it fascinating that it sounds like your school is intentionally creating the language needed for your data warehouse, before you're actually even collecting it, most likely. And so what I'm curious about is, how do you maintain that level of governance that is required for really sound data research to take place?
Hudson Harper 15:44
That's such a good question, what you mention is a lot of intentionality, and also a lot of very clear documentation, in all of our conversations, of our different data sources. There's so much that we could do, it would be counterproductive, and maybe sometimes feel like we're drinking out of a firehose, you know, if we weren't just focusing in on what are those key things that matter to us. And you're absolutely right, having that in place made our transition over to utilizing a data warehouse that much simpler, because we sort of had everything mapped out. We just made our data warehouse an exact mirror of that. And now, our data warehouse is helping to guide even further institutional research and data governance because of the various structures like the fact that you can define schemas within your data warehouse, right, kind of defines, how is that data going to be maintained?
Bill Stites 16:39
Before we get in kind of like the data warehouse side of things, one of the things that we spent some time working on and I know that you've really done a good job at documenting where, at least in terms of what we put together in the guide, are all of the different types of roles that people are going to have at a school around data. Given the different sizes of schools, you know, you could have one person in each one of those roles, or you could have one person wearing many hats in a school that do multiple things within the context of that. Can you enlighten people who may not have read through the guide what you see those roles as being in a school? And then how, by example, that might look at your school or how you might see that looking at others?
Hudson Harper 17:27
Yeah, absolutely. And, you know, I highly recommend that if you're even remotely interested in institutional research, pick up a free copy of our guidebook – you can find it on the CIRIS website hosted by Maret.
And, you know, as Bill was mentioning, in that guidebook, we lay out sort of like this hierarchy within your data governance committee, where there should be like one person who is in charge of overseeing the mission, and sort of like the day to day operations of that data governance committee, sort of like your data governance czar, who's going to say like, “Okay, we're going to pull together all the people across the school who touch our data, we're going to figure out what are their needs, what currently exists. And then we're going to figure out a strategic plan to how we can sort of wrangle that all together.”
And then below that, we have people who are data stewards, so people who are actually working with different departments and working with different databases, who can have a sense of ownership over that data to make sure that that data is being properly maintained, but also to understand sort of the value and purpose of that data.
Then below that we have data custodians – people who are actually doing the data entry, people making sure that the quality and integrity of that data is maintained. There's a common phrase in data that “garbage in equals garbage out”. And certainly, we don't want to be working with garbage in our system. So it's really the custodians who are making sure that our data is properly maintained.
And then, I think this is sometimes overlooked, but super important, there is the user, right the consumers of data, who should be part of this governance committee, because they're the ones who are going to enlighten everyone else on what is the actual purpose and value of that data.
Like, if you're a teacher, right, you probably have better insights on what student data is relevant to supporting that student on a day to day basis than someone who's looking at this from 30,000 feet above.
At a school like ours, with only 15 people, you're absolutely right, Bill, there's a lot of overlap. In fact, you know, I think, Christina in an earlier podcast, you used the phrase “data cop”, and you asked the question, “are there any schools where there's just a singular data cop managing everything?” And I hate to say it, but, I might have been that data cop during the first couple of years of the school, because it took a while to develop the structures and organizational tools that we needed to improve our data governance, you know? And now today, thankfully, I think there's a number of us who are wearing multiple hats within that data governance committee.
So it's inevitable that I end up touching almost every aspect of our data except for sensitive or private information for individual students, right, but I'm seeing sort of like, what are our data governance goals? What are our different systems? How does that data get maintained, cleaned, properly handled?
But there are other people like our data operations manager – she's like our primary data custodian. And we've tried to bring in multiple people who I think typically, you'd find in other schools, like our director of college counseling or director of student support, right, they touch so much data, of course, we want to make sure that they're in the fold of how that data is maintained.
Hiram Cuevas 20:33
So Hudson it reminds me of two other areas of campus, when you were talking about the data cops, you know. In admissions, you often hear that everybody is a member of the admissions department. And you also hear that everyone is a member of the advancement department, in terms of fundraising. I do like the fact that you're suggesting that everybody is responsible for the data, and everybody needs to have some sort of responsibility to the integrity of the data. Now to help our listeners, could you define for us the concept of the data warehouse/lake? And what would that mean, in terms of how it differs, say, from general systems that we use?
Hudson Harper 21:14
Absolutely. I mean, so data warehouses and data lakes are different structures for storing data from multiple sources. And I know those two words get said together so often, it's really easy to confuse them. And in fact, in reality, what most people are using when they talk about either of these is a data lake house. I know like the vernacular is so great, right? There's no confusion here at all.
But typically, a data lake is a place to store unstructured data. And when I say unstructured, I mean things that don't easily fall into, say like a table, or things that you wouldn't normally see in like a spreadsheet. So you could think about things that fall into like PDF files, or maybe these are old surveys, that people filled out on paper and had been scanned, or, you know, if your school didn't just recycle all the parochial slips, right, maybe you have an archive of those.
And these are typically really hard to work with, just because of the fact that they're unstructured. I will say there's been a lot of advancements in the last couple of years thanks to AI and being able to extract information from those unstructured data points.
But when it comes to working with structured data, that's where your data warehouse comes in. And so that's the idea of, we have all these underlying tables of data from either your SIS, your LMS, whether it's from Google forms that you use for surveys, right – all those tables of data can be stored within your data warehouse. And the point of that is to easily be able to connect those different data points. I mean, it's really hard to say like… Okay, we give out a student wellbeing survey every quarter of The Downtown School, and we ask questions like, hey, how much sleep are you getting, on average per night? And we want to relate that to, say, how they're doing on their average day to day assignments in Canvas? Well, how do you get those two data points to talk to each other? How do you relate those to each other? And the data warehouse is the place where that can happen.
Bill Stites 23:11
It's interesting, because, you know, Hiram and I have often had conversations with other schools where they're like, “Okay, what's the one database that we can use that's going to do all these things?” And we will often say that the idea of the single database is a complete and total myth. I think what you're talking about is really a great example of that, whether it's a structured warehouse or it's the lake, or I love the “lake house”, I think that's brilliant. But this idea of, you know, we've got all of these different data sources.
I mean, one of the things I think people on the podcast probably know, I know, everyone on this call knows this – I love a good data map. I love laying those things out and putting them together. So when you see all of those things, how do you get all of those pieces to talk? And I think what you're describing, you know, that structure that you have, from that visual down to getting at those pieces can really be helpful.
I also think that when you start talking about the ways in which schools move from system to system, there's the need to hold on to legacy information or information that you might need when you're going from system A to system B, how much of that you bring over? How much of that do you hold on to?
So I think when you start thinking about it from a long term governance standpoint, outside of data governance, but good school governance, having a good warehouse where you’ve got all this information in both of those forms, I think plays out very, very well.
One of the things that I want to kind of pivot to is something that you said, and something that I've seen when we got together this summer, you kind of demoed lightly for us, and I know you've talked more about it, and I've been doing more work on that, is where do you see AI coming into the work that we're doing when we're trying to make sense of all of this data? And how can it really help?
And I think when I ask that question, I want to preface it by saying, my background, you know, early childhood Ed, third grade teacher, thought I was going to be an art teacher. And even my family would always say, “explain it to me like I'm a third grader.” Not meaning talk down to me, but just simplify it to a point where I can understand if I'm that third grader, I'm gonna know exactly how AI is going to help me solve these types of problems.
Hudson Harper 25:27
All right. So when you told me to prepare for this, in the notes it says, “Don't be afraid to give hot takes”, basically. I have some serious hot takes when it comes to the use of AI for this sort of stuff.
Christina Lewellen 25:39
And we're here for it. So you've come to the right place – bring it on.
Hudson Harper 25:43
That’s great. Let me start with what I think are the best and most benign uses of AI that we can start using today.
So, Bill, you mentioned that there's a lot of technical work that has to happen to get these systems to talk to each other. Unfortunately, because a lot of these vendors and companies like to think that they are the one database solution for all of our needs, they haven't necessarily invested in all the tools that you need, to make them talk to each other.
I think there's been a lot of really great work done recently to try to push for universal standards between these systems. But they're currently not there.
And so one thing that AI can help with is on the technical side – it can be your sort of copilot to help you develop the technical pieces between the systems, say, if you're trying to write your own code to utilize an API to connect two different systems.
One of the things that I've been doing a lot over the last year, year and a half, is writing my own Python repositories to help connect things like Blackbaud or Veracross to your data warehouse. And, you know, I would like to say I can take 100% ownership of that code. And ultimately I do, but, AI certainly helped me write those connections. And the reason I think those are pretty benign is the AI at that point isn't helping to do any of the analysis. It's not helping to pull any the data together directly. It's helping me do that job.
Because one of the things that I think is super important, and again, going back to an earlier podcast, Christina, you said that last year, everyone was talking about cybersecurity – that was the hot topic. And this year AI is the hot topic. And here I am thinking please, please let cybersecurity be the topic alongside AI, because I think there's a lot of unexplored risks and considerations that we still haven't fully unearthed when it comes to utilizing AI with this type of data.
That's not to say that we can't. In fact, I think that's where the like the next level of AI use comes in. Right, there are secure ways to have AI do things like write code behind the scenes to analyze and pull out trends in your data.
I remember we were all sitting around the table at CIRIS, and I was like, hey, are y'all curious about this AI agent that allows you to put in a CSV file and then it will start running and thinking about code and will give you the results, create charts, run statistical tests. And I think that is a glimpse at what is going to eventually be possible.
The problem that you run into is I had 100% faith that that was going to be secure because I was writing my own code to access AI through, you know, we were using Open AI’s API, which by the way, is not a free or cheap thing, necessarily. But I knew that I had taken the appropriate security measures to make sure that that data wasn't going to be leaked anywhere else.
Bill Stites 28:34
I think that's one of the biggest things that I know we are struggling with here at MKA. I know that it's one of the things that comes up often in some of the workshops and sessions that ATLIS has run, in terms of taking actual school data and putting it into AI and having it analyze that and understanding what the exposure is there. Is it going to be used for training the model like is it going to leak somewhere? How do you get around all of those things?
And I think, having seen what you did during that session that you were talking about, the other piece of that is, you know, one of the roads I went down myself – a long road I went down – was trying to figure out how to make an API work to do exactly what you're saying, knowing no code whatsoever, but having at least a systems understanding in my head.
But then I think about the other people that were around that table with us. I think about the other people that are at all of our schools. And maybe around that table there may have been me, you, and maybe two others that may have gotten the code aspect to some level of what you were showing, whereas the others don't have that.
So where do you see that shift happening between the work that you're doing and you're spending to write this code, to get it to what I'll go back to is that third grade level where people in the development office that aren't coding, the people in the admissions office that aren't coding, the people in the academic office aren't writing the code to do all this but they need the information to know that it's secure once it goes in, but once it goes in, how to get those insights out of it? Where do you see that happening and how?
Hudson Harper 30:22
I think we're already seeing that sort of happen. With as much out there about using Chat GPT, or using Claude or any of these other large, AI platforms for your work, you know, people have already started using those to do things like data analysis. You do have to pay for Open AI's premium version of Chat GPT, but honestly, the data analysis plugin is pretty good. And as long as you have completely opted out of letting Open AI use your prompts for training purposes, I feel like it's reasonably safe compared to, you know, if you just put student data into the free version of Chat GPT, please don't do that, anyone listening.
Christina Lewellen 31:05
Can we pause there for a second, for people who aren't quite as familiar with this and don't have the background that you guys have? What you're saying is, if someone wanted to use their students’ data, or their school's data, and have AI help analyze it, you're saying that as long as you're opting out of, like, please don't train the black box of AI, at this point you feel like it's relatively safe and that that's how people should be using it.
I think a lot of people are still scared. I mean, I just want to like, stop down on that point. Because I think as I talk to audiences, the big concern is, but if I put my data in, where does it go? The black box question is still a big question. So can you just like say that, again, to reassure me that that's indeed what you're recommending.
Hiram Cuevas 31:52
And I'd like to add, because you're talking about the API as well, and so, I've also heard from others say that if you're using the API, that it is not used. That data is not used for learning purposes of the black box of AI. And I'm curious about that switch that you're referring to, about, do not use my data, because then I also hear that Claude AI is a much better approach for some of these issues that we're talking about.
Hudson Harper 32:19
All right, and thank you for pausing, because I definitely don't want to go out on the record and saying, yes, please start throwing all of your student data into Open AI.
Look, I think there's still so much that's not known. I'm just saying in terms of what people are doing right now. Sort of what are the safest ways to approach it? And certainly, Hiram, thank you for pointing out, like, using the API directly with some amount of coding, is the safest way. It's still not great, in terms of are you fully aware of where that data is being stored?
So for instance, that data is still being stored on Open AI’s servers for 30 days, unless you happen to have a very special agreement where they have a zero-day retention policy for whatever you feed that API. But it's still something to consider. You're putting student data onto someone else's system, that, while they maintain that they're FERPA COPPA compliant, have you really entered into an agreement with them to guarantee that it is being securely handled?
Christina Lewellen 33:20
So where's the line for you, Hudson? Like, as you're thinking about this, and you're a little bit more advanced, you feel a little bit more comfortable with your ability to code and have these protections in place, but where's the gut check? Like, what's too much for you? Where do you go, “Oh, no, I'm not doing that.”
Hudson Harper 33:35
I’ll start by saying I am enthusiastic about where AI is going. I don't want this to come across as a bunch of FUD – a bunch of fear, uncertainty and doubt, for the listeners. But for me, I don't know if I'm necessarily comfortable to put student data into one of these big AI platforms. And I think there are ways to do it more securely. Again, like for instance, if you do all the anonymization by hand, before you enter that data into Open AI, or if you use tools to create what's called synthetic data from your datasets, where you're using things like the specialized local models to anonymize and then enhance your data so that you're not putting in like the raw values of like, you know, this is a student's name, gender, race, etc. Those are better ways to use things like Chat GPT, to analyze your datasets.
But for me, I think where I get especially nervous is with the rush of all these different AI tools and companies emerging saying that “we can solve all your problems through your data by just plugging into our AI magic boxes.” And you know, it's been very fascinating to me to like to see what's emerged. And one of the things that we've done with our professional development at The Downtown School is train faculty on how to read things like the data privacy policies, and to look for things like, hey, even though we're using Open AI's API, behind the scenes, we may use some of this data to train our own specialized models to look for that sort of thing.
For me, one of the wildest examples I've seen of this is an AI tool…I won't name names that are not ready for that kind of hot take.
Christina Lewellen 35:16
Oh, come on, chicken. [laughs]
Hudson Harper 35:20
But saying things like, we can help you with student IEPs. And I'm like, please, no, we're not ready for that.
These companies themselves are still trying to figure out what does responsible AI usage look like on their end. I don't know if we can necessarily trust them at this point.
Christina Lewellen 35:37
All right, it sounds like we found your line and it's – IEPs. So that's good to know.
When you're looking at the work that you've done, because you definitely have your brain wrapped around this, I think more than many, is there a particular project or a particular journey that you took in terms of data analysis that you're particularly proud of? Whether you leveraged AI or not, as you talk about creating these systems and creating the governance for it, what has been a really successful case study that you would share if a colleague called you right now and said, “Hey, I don't know where to start, tell me something cool you've done.” What would you say is, kind of, your favorite data journey?
Hudson Harper 36:18
Aw that’s such a good question. And I would like to say that there are so many neat things that could go down. But for me, like, my favorite projects are those that not only help us institutionally, but really, help out the individual teachers.
And so one of the things that we've been doing since the beginning of the school is think about the ways that we keep track of our feedback and comments that we give to students. A lot of schools do narrative comments for their students at each quarter. But I think it's something that we've done particularly well and have found ways to meaningfully communicate real actionable feedback to individual students.
And that's a lot of data that gets pulled together over the years. And the way that we have been able to maintain and keep track of that. And then more recently, something I've been super excited about is to take that data and then warehouse it in a way that all of our teachers can go back and access. It's just this treasure trove of the lived experience, or at least the curricular experience, of students at The Downtown School.
And one of my favorite personal projects around this was I was super excited and honored to be picked to be one of our graduation speakers last year. I was like, this is a great opportunity to expose both the highlights, and I don't want to say low lights, but you know, some of the more interesting failures, from our students by pulling all this data together about them. And some quick figures, you know, we were able to pull 70,000 different unique comments given to students over the course of four years.
Christina Lewellen 37:50
Well I think it's interesting that you're bringing up that you have “institutional researched” your way into a commencement speech, is that what you're saying? I think that's kind of a unique use of data. I bet not many people thought to do that.
Hudson Harper 38:04
Yeah. And I was like, I know this is playing into so many stereotypes about me as the math and tech person, but I was just like, it was too good to resist.
Christina Lewellen 38:13
You're in a safe space here, so you can share those little secrets, because I'm sure that that was cool. I'm sure the outcome was really neat, to actually pull on data that is existing and real, and to use that as you kind of summarize the journey of school. That's really cool. I like it.
Hudson Harper 38:28
Thanks. I mean, but beyond just sort of that fun side project, the dividends that's paying today is when teachers go to write their quarterly comments, they can see everything that they said to a student at a glance, at something that is pretty hard to do in your traditional learning management system. And where we're kind of headed, though I don't think fully, is using AI to take all this work and effort that teachers have put into giving us feedback, and to look for those commonalities, those trends, and the feedback that they're giving, and also to help summarize that feedback in a meaningful way for students.
Hiram Cuevas 39:04
So Hudson you bring up an interesting next step. So you've got this data warehouse where you've got all of these comments, and I'm sure a lot of schools have lots of data in their respective SIS – what's the tool then from within the data warehouse so that each faculty member now has access to that repository that you've now developed? Which, from an SIS perspective, oftentimes that's limited by roles and permissions and whatnot?
Hudson Harper 39:31
That is a great question, because I think one of the big benefits of having a data warehouse is it makes managing roles and permissions a lot easier than if you're trying to just dole out CSV files of all this data to everyone individually. And so like, for example, with those comments, they all live within our data warehouse, but those connect directly into a Google Sheet where we can set up individual permissions for different filters or queries so that when a teacher pulls up that spreadsheet in Google Sheets, they only see their feedback in comments.
Bill Stites 40:05
One of the things I wanted to ask, and it's kind of to go back just a little bit, is when we were talking about the data warehouse and the data lake and what you're storing in each of those. And what we're seem to be focused on with our conversation around AI, is how well AI can help us with that structured stuff, everything that's going to be in our warehouse.
Have you come across a tool, or an example, or anything where AI has been able to help with that unstructured data sources? Whether it's a PDF, you know, whatever it may be? Have you found anything in that area yet, or is that still to come in your mind?
Hudson Harper 40:44
So I think there's so much potential there, particularly in terms of extracting how your data points from a PDF file.
For us, this isn't as much in our minds, because we don't have a lot of archived data that we need to be extracting from. But something I think about quite often that I want us to use more is using our unstructured data and using AI to enhance our processes.
So there's sort of a technique in AI called retrieval augmented generation, which allows you to feed things like PDFs that might contain things like policies, handbooks, maybe all your documentation for how your school operates. And the AI can help you to search through and find relevant information about, say, hey, what do I do, you know, if a student, for example, I don't know why this came to my mind, but, hey, what if a student accidentally eats some peanuts? Like, what do I do in this situation? AI might be able to help you look through all that documentation and tell you just in plain language, “Hey, this is what our policy says.”
And I think that's something that, of course, not just schools, but all organizations are going to be using AI for. But it's going to be that person that you go to, and you have like, gosh, I feel like this is a really dumb question, please help me out. AI is going to be able to help you out with that stuff.
Christina Lewellen 42:04
I know that you are slated to present at NAIS on your distributed leadership ecosystem. And we talk a lot at ATLIS about leadership roles that tech leaders can kind of step into, because most tech directors don't necessarily expect to find themselves in school administration roles. And you have a unique situation at your school where all of the educators are in leadership roles. So that's kind of cool.
But if you don't mind, I'd love to ask if you have any thoughts or advice for tech leaders, who maybe don't sign up for something like The Downtown School, but they find themselves in these administrative leadership positions? That can be a big shift sometimes, and I think some of our tech leaders can find that a bit intimidating. So do you have any thoughts or advice that you would share with tech leaders, or, you know, unsuspecting math teachers that find themselves in tech leader roles, and kind of how to handle that step into a leadership capacity?
Hudson Harper 43:10
Yeah, absolutely. I mean, for me, one of the big differences between The Downtown School and your traditional school with a very set organizational chart is that we have a lot of dotted lines between our various people in roles. And I think even within a traditional organization, it's about how do you find the sort of unofficial dotted connections with people from multiple departments?
I think, especially in data, we talk a lot about data silos, and a lot of that comes just because each department has their own operations, they have their own process; oftentimes, they have their own system. And to me, I look for how can we break down the silos? How can we form connections and collaborations between them? And for me, that's one of the benefits of being at my school, is those connections happen so naturally, but I think they can happen, as long as you're intentional about it, anywhere?
Christina Lewellen 44:07
It's interesting that you say that, because I hadn't really thought about the fact that in a lot of ways, technology directors, if they're in a more traditional tech director role, a lot of the work they do is breaking down silos. And so, in a weird way, they're very well suited to leadership roles at their school, because they're just so well versed in those, kind of busting down the silos mentality.
So I hadn't really thought about it till you said that, but I can see… Are you excited about presenting this concept to NAIS? I'm sure it's going to be well received. It's very different. Your school's unique.
Hudson Harper 44:39
Oh, my gosh, I am excited, because I think that we have a lot to offer in terms of just a different way to think about the school model.
We're going through such a transitionary time and independent schools, that... I think as schools need to adapt to these new circumstances, they just need to broaden their toolkits, broaden the different ways to think about organizing… Not necessarily even the whole school but, our goal at this presentation is to show you, within your department maybe, how can you implement some of these ideas to be a little bit more nimble, to be a little bit more, you know, prepared for the future?
Bill Stites 45:14
So Hudson, one of the questions that I think we want to get through with all this, you know, AI seems to be this conversation of what's coming in the future. What do you see, as the next thing on the horizon for us when we start talking about these things? What should we kind of be paying attention to or looking for?
Hudson Harper 45:33
I think, for us, it's about continuing to learn how to adapt, and to evaluate these AI tools. We have a particularly eager faculty here at The Downtown School who are already thinking about all these different use cases. And I think, honestly, like, if you sort of start buying into like the idea that AI is going to be a saying and that we need to learn it, I think there's so much content out there about different use cases for things like pedagogy that, like it's not hard to find.
I think what, again, we need to focus on is how do we use it responsibly, and how do we keep, on top of all the downwind implications of AI on things like security, how do we keep track of its impact on our students’ well being – that's what I'm, I want to focus on.
Beyond that, there's some exciting things happening sort of in the CIRIS community that I want to make sure people are aware of. I don't think this has been officially announced so sorry, Eric, if I'm jumping the gun here, Eric Heilman, the director of CIRIS, but we're gonna be doing a professional learning group next semester in the spring, dedicated to AI applications for data and analysis. That's
Christina Lewellen 46:40
That’s incredible. We're gonna definitely break that news on this podcast. That's awesome.
Hiram Cuevas 46:45
That's fabulous. I bet there are gonna be a lot of people interested in that one.
Hudson Harper 46:49
Yeah I’m super excited. And I promise it won't be me just saying like, “don't use AI because of security risks,” right? I promise to give you some actually useful tools and use cases for it. Beyond that, looking forward to hopefully being at ATLIS and participating, you know, spreading the gospel of data warehouses, different use cases.
Christina Lewellen 47:10
This episode has definitely been an opportunity for you to kind of grease those skids and create some interest in the ATLIS conference. And obviously, everything that you are contributing and participating in the CIRIS world, we really will look forward to keeping our finger on that pulse. And, you know, you're welcome to keep us posted. Come back and chat with us anytime.
As we sort of begin to wrap up this episode, which has been so incredible, I give you a lot of credit for kind of being on the leading edge of these conversations, because there's not a lot to draw on. I mean, this is very new territory. So I give you so much credit for being brave and kind of venturing out into this space. I know that your school is just perfectly suited for it. So this was really wonderful conversation.
But before I let you go, I started this podcast today by asking my co-hosts about their soul city, and we learned something new about each other. Can you tell us what your soul city is? Where do you love to visit? Do you have a favorite city that speaks to your heart?
Hudson Harper 48:13
Oh my gosh, the irony that you all said “Seattle”, and here I am going to say it's actually Boston for me.
Christina Lewellen 48:18
Boston's a good city.
Hiram Cuevas 48:21
Boston’s my number two.
Hudson Harper 48:22
Yeah I mean, I grew up in the South, but I think some part of me transitioned over to being a Mass-hole when I was up in Boston. You know, give me a good ol’ lobster roll, park myself in Harvard Square, like, I love it.
Christina Lewellen 48:35
Okay, we're gonna go with it. I mean, we can't all be Seattle, and you're there every day. I think that if you had said Seattle, I would have called you a liar, so, I'm gonna go with Boston. And I'm just saying that from this conversation, I don't see very much Masshole in you. But if you want to claim that, we'll go ahead and let you claim.
Hudson Harper 48:52
I'm not behind the wheel right now, so...
Christina Lewellen 48:56
Oh, okay. Got it. Alright, so we're not driving together ever. Mental note. Noted. Got it.
Hiram Cuevas 49:00
Well, you got the New Yorker and the Bostonian, look out.
Bill Stites 49:04
You can bring Philly into the mix. Let's go.
Christina Lewellen 49:07
We're quite a tri-state area.
Hudson, thank you so much for being on Talking Technology with ATLIS. This was a wonderful podcast episode, and we really look forward to following your work. I have no doubt it's going to be pretty exciting stuff.
Hudson Harper 49:19
Yeah, absolutely. Thanks for having me. It's been a blast.
Hiram Cuevas 49:21
Thank you so much.
Bill Stites 49:23
To put in a big plug for Hudson’s website, I know he does a very nice job at documenting and sharing out everything that he's working on, so I just want to thank him for that. And again, put in a plug for his Harpertech.io website, which we'll be sure to put in the show notes.
Hudson Harper 49:40
Thanks, Bill.
Christina Lewellen 49:41
Thank you, everyone for joining us. This is another episode of Talking Technology with ATLIS. We'll see you next time. Have a great day.
Narrator 49:49
This has been talking technology with ATLIS produced by the Association of Technology Leaders in Independent Schools. For more information about ATLIS and ATLIS membership, please visit theatlis.org. If you enjoyed this discussion, please subscribe, leave a review, and share this podcast with your colleagues in the independent school community. Thank you for listening.