Season 2 | Episode 2 | Dr. Adam Ross Nelson, University of Wisconsin, Madison
Colleges and universities face a new reality. They are being held more accountable than ever to prove and improve the value of an education from their institution. As more low-cost degree programs, microcredentials and digital certifications emerge, the more choice learners have. Institutions must shift to an outcome-based strategy to meet the needs of the modern workforce.
How? Build a sustainable culture of data at your institution. Dr. Adam Ross Nelson, instructor in statistics at the University of Wisconsin, Madison, shares his step-by-step strategy.
[0:06] ERIN KING
More people than ever are questioning the value of higher education. We're here to explore why they're right, why they're wrong and which institutions and individuals are rising to the challenge. I'm here with our Analytics Consultant, Dr. Jacob Bonne.
So what do you have for us today, Dr. Bonne?
[0:24] DR. JACOB BONNE
Today we are here with Dr. Adam Ross Nelson, Data Scientist and an instructor in statistics at the University of Wisconsin, Madison. Dr. Ross Nelson has also worked as a contractor for the National Center for Education Statistics, among other clients.
He works in career coaching, helping other mid- and late- career professionals transition into data science careers. Dr. Nelson, please share with us a bit about your professional experience and your passion for data in higher ed.
[00:49] DR. ADAM ROSS NELSON
Thank you so much for having me today. I am so happy to be here. I'm so happy to have a chance to look at the value of higher education with all of ya'll. I would say, and this is a spoiler alert for anyone who really digs into the research on this, that higher education is worth it. And this is a podcast, you can't see me doing my air quotes, but if you look into the empirical research on this, higher education does have positive returns on investment.
More about me: I started my career in education, and I've spent most of my career in education. For example, I started my career as a teacher of English as a foreign language in the late 1990s. Then, much later in my career, much more recently, I teched up, transitioned into data science, and my main role now as an independent consultant is to do data science for hire and then also provide career coaching to other mid- late-career professionals who are looking to transition into data science
So you posted something on LinkedIn a few days ago that said, "Is anybody else sick of gatekeepers?" So I want to know what sort of gatekeepers exist in higher ed when it comes to data, and how would you advise a faculty member trying to overcome these limitations?
Oh, that's interesting...you know, I worked in higher education a lot myself, and I've been rethinking about how I spend my career in higher education; and in a cynical way, (this is overly cynical) but on some views, you could think of folks who work in higher education as professional gatekeepers.
I put some emphasis there not to oversell it, but to really force folks who can hear this, who are in reach of my voice, to think about: If you work in higher education, are you really serving as a professional gatekeeper? And I know there are points in my career where I was doing my job. I was doing what my boss wanted me to do. I was doing what my colleagues wanted me to do. I was doing what the institution hired me to do...I worked at state institutions with [what] the Legislature wanted, [what] the public policy weighed in favor of, et cetera, et cetera, et cetera.
I was clearly doing my job. I was meeting my expectations, but I was really keeping gates. And the Admissions process, for example (I've spent a lot of time working in Admissions) so the admissions process isn't just one gate. It's many gates, and every question that a school asks in the Admissions process is another gate.
For example, many schools have recently patted themselves on the back for going test optional, and this is a step in a very good direction. But some of those schools are also asking students who elect or select the test optional process or the test optional options, they're asking this supplemental question: Please explain your reason for choosing the test optional process. This is, and I just pause, because I have to think to myself, why is this additional explanation on the test optional process necessary? And if you ask for an explanation, are you undermining the optionalness?
I think the answer to those questions are that it's probably not necessary if you're truly in the test optional mindset and also asking those questions for reasons related to the test optional experience, again, undermine the optionalness.
In Admissions, you have to remember anything that a college or university asks in the Admissions process, even if they label it optional, it's not really optional, because the student sees that as an opportunity to communicate with you. And the social forces at work and the institutional forces at work really motivate students to answer even optional questions.
And if you're thinking about gatekeepers as they relate to data, and also, how would I advise faculty members to overcome the limitations to data, there are all kinds of limitations and gates associated with data. And then, of course, which faculty member are you? Are you the faculty member in math who's trying to help the student succeed in first semester math? Or are you the faculty member in education policy who's looking to collect a large data set from a college or university in order to do research and support your publishing and scholarly agenda?
I think the answer for those two faculty members are...very different. So in generalities, I say it's about building data culture at the institution, and it's about building a shared sense of responsibility regarding the data. And it's about building a shared sense of responsibility for using that data and being good stewards of that data for the purposes of promoting student success, student access, student completion and efficiency, etc.
You absolutely bring up several great points there.
One you've touched on is data culture, and so talk to us maybe a little bit more about that. Leveraging data across higher ed has always been an interesting component of of my higher ed journey and higher ed career.
Folks are thinking about data in different ways, leveraging data in different ways. So what are your thoughts? Where are some opportunities there for higher ed, and how could this data culture grow and evolve?
Absolutely one of my favorite all-time ever favorite questions.
I like to go bookish on this, and I start with: What is culture? So I think maybe...some of my colleagues in education might appreciate the stepping back and breaking this out just a little bit and really focus on, again, what is culture?
So a frequent definition, this is my definition off the top of my mind, it's a set of shared values, traditions, rituals, language, words, actual words and other customs that pass from one generation to the next.
So if an organization can take it upon themselves to think about how they interact with data in terms of their values, their traditions, their rituals, their language and other customs, that is one way you can begin the process of identifying the culture in specific, tangible and concrete ways. (The data culture in specific, tangible and concrete ways.)
I do a fair bit of corporate training in my data science work, and in my training sessions I always urge organizational members to have conversations about data. Start with the rudiments, and when you're having these conversations, you have to know what to talk about. So I always give model questions that you can ask each other in either formal or informal ways.
You can do this in passing, or you can do it in a very organized way. But some of the rudimentary questions are: What is a data set for our purposes? What is data? What does data mean to us? What does it mean for us to do an analysis?
So let's say, if you really think about the process of doing an analysis, ask yourself, what is that analysis? What are the inputs? What are the outputs? Who's involved? How are they involved? When are they involved? And then as organizations start to have these discussions, I also say document those results. Write down the results of those conversations.
And this is the process, by the way, the process of having those conversations, those conversations then themselves become a tradition, if you can repeat them. And then, of course, pass them on from one generation to the next. The act of having those conversations is a tradition. It's a ritual. And then having the conversations forced you to develop the language, the shared (I wish you could see me doing my air quotes here. I'm having an air quote fiesta) in the shared language and data-related vocabulary through the course of having those conversations.
One of the best organizations who really, really [has] strong data culture, in the equivalent of their staff manual they have a glossary; and they're a very data-centered organization in general. But this glossary defines these words and what it means specifically for that organization. [It's a] major asset for that organization's data culture and whether that was intentional on their part (and actually now I want to go ask) was that intentional on their part or not? I don't know. But they obviously had the conversations. They thought about what those words mean for them.
And I have a couple of articles on this topic. One, for example, explains, what is a data set? And I usually share this article with folks who are interested in starting these conversations. And I'll say, here's my definition of a data set. Here's the article I wrote about it. Use this as a starting point for your organization. (It's a short article.) Have you and your colleagues read that article and then sit down and discuss: OK, what resonates with you? What do you accept? What do you reject? What works for you? What doesn't work for you, and then decide what a data set is for you and your organization and your team.
This will, I guarantee, this will boost and build your data culture. So I try to be really pragmatic. I'm both bookish and pragmatic when it comes to talking about and promoting data culture.
Yeah, absolutely. I love that idea of balancing the pragmatism and the research or bookish brain, as you mentioned there. In my conversations with folks, often we talk about leveraging, how can we leverage and bridge the gap between those two pieces. I often remember telling folks, you've got to start with a research question before you can dive into data sometimes.
And we think about those things from our grad school days, but don't always continue to perpetuate those best practices for research or for publication.
I agree and on the conversation to build on what you just said, if you are an organization, you can think about, OK, so what makes a good research question for us? And another really good technique that I tell organizations to do is start building a backlog of research questions and analytical questions, because you can't study or analyze everything you want to right now, you have to prioritize. Some things you have to save for later. Some things you have to outsource, et cetera.
So building that backlog, then, is another data related cultural artifact that can really move your data culture forward.
Something else we wanted to touch on was another article that you published for AIR on standardizing IPEDS data at institutions, which really is a great step-by-step guide for leveraging some of the publicly available data in higher ed.
What are some applications you see, even perhaps on that pragmatic side, where researchers might use this data or our higher ed researchers as well?
There's really probably two reasons you can use Integrated Post-Secondary Education Data System data and IPEDS for short. If you're not familiar with what IPEDS is, the first way is to conduct institutional research. Well, what's institutional research? It's a lot of things, but one of the things that institutional research professionals do is they'll compare institutions.
They'll compare Institution A with a set of institutions called the Comparison Group, or they'll just compare Institution A with Institution B. Well, to make those comparisons, you need reliable points of comparison, reliable data. Well, you can get this from IPEDS.
And anecdotally, seasonally (this is an example why this is such a good article if you're interested in comparing organizations). Seasonally, you'll be at a professional conference, and maybe you've experienced this, too. You'll be at a conference where a presenter says that they're sharing data on a group of institutions, and they describe in their methods that they went to institutional websites to get the data for their work. They just combed through the institutional website, the .edu. And then...you'll hear the presenter say something like, "I had a student go to the sites and get the data." And they are literally transcribing it from the website to a spreadsheet.
Well, as it turns out, the US Department of Education, a great organization, obviously, tabulates and provides this data in a CSV format, and you can just download it. So that's the first use. Just simple comparisons to learn about organizations to compare organizations in a rigorous way.
And one of the nice reasons about using IPEDS data rather than relying on information from a website is sometimes the information on the website is: Is this last year's data? Is this next year's data, or which year did it belong to? But when you get it in the CSV format from IPEDS, you know exactly what year it's coming from.
And another very common use for these data are research. So for academics, academics will use the data in their research and publication agenda, and to find examples of this is very easy to do.
...You could go to Google Scholar, for example, and keyword search "integrated post-secondary education data system." And then the articles who reference this IPEDS data in their methods section will pop up. You could do some other limiting factors as well to refine that.
But if you're looking for examples of how researchers use IPEDS data or other similar data, such as College Scorecard data, you can find that fairly easily yourself as well.
Dr. Nelson, I wanted to bring up some career advice that you share on your LinkedIn profile, which is a three-step process for becoming a data scientist.
This obviously doesn't happen overnight. We actually had an interview with Dr. Jennifer Priestley, who's a Professor of Data Science and Analytics at Kennesaw State, and she warned against the one-weekend career changers.
So what did the career path to becoming a data scientist look like for you, and how fast can someone realistically expect to become a data scientist that top employers want to hire?
It's a tough question, because the answer is really specific to the individual, and that might sound like a punt, but I promise you, it's not. Really every person, especially for mid- and late-career professionals, one of the reasons I focus on working with mid- and late-career professionals...I think the answer is more systematic, a little bit more straightforward for earlier career folk, but for late career folk, the answer is so specific, because by the time you're in the mid- or late-portion of your career, there is so much diversity in how you got there.
And also for mid- and late-career professionals looking to make that transition, it's highly contingent upon your career. And in fact, one of the main strategies is to rely on the previous experiences you've had in other roles in order to really bring value as a domain specialist to data science.
I recently had a mother of a prospective client contact me (not kidding, I'm not kidding about this). But the mom said her son didn't finish college but always aced math, did really well in math, and I believe it. And the question for me was, could her son become a data scientist? And I said, well, my first response was, I'm not the gatekeeper, echoes one of the questions we were talking about earlier. I am not in the business of being a gatekeeper. I was in that business for some time, but I'm no longer in that business, so I'm not in a position to say to anybody, don't pursue a career in data science.
I replied that there are paths to become a data scientist. And then explained if her son has a career history on which we can rely, I might be able to help find a path for that transition.
But I also said I'm not in the business of inflating anyone's expectations. The process is very hard to become a data scientist for most folk. And on that point, if you're thinking about making a transition into data science for the mere sake of making a transition, it is a mistake to pursue data science.
A short tour over at the Bureau of Labor Statistics will reveal that there are other very high-paying career choices with lower barriers to entry. Maybe this is homework for people listening to the podcast. If you're thinking about data science, and you're not sure about data science, go over to the Bureau of Labor Statistics, and look at the career prospects for other careers not related to data, and you will see faster growing fields that pay as well with lower barriers to entry for data science.
So the real question for the mom's son and for anyone looking to make the transition into data science (I agree with Dr. Jennifer Priestly. I do. I agree that one-weekend career changes [are] probably unrealistic), so the real question is: Is the juice worth the squeeze? Do you have the time and the energy and other resources to begin and then complete that transition?
So spending all the time, energy and resources, that's the squeeze. And then you have to look at what data science offers you, what you have to offer data science and then decide if you're going to make that transition.
So for example, I was doing some writing on this topic very recently, and the first chapter is about who you are and what you are looking to do with your career. And it's not about the challenges associated with becoming a data scientist, and it's not about the perks and the rewards associated with becoming a data scientist. It's about getting clarity on who you are and what you're looking to do with your career.
So if you answer those questions related to who you are and what you're looking to do with your career and they point you in the direction of data science, I go back to one of my earlier refrains. I would say, go for it. You make the plan and set reasonable expectations, but go for it.
In addition to that, we'd love to get your thought if we scale that topic back a little bit more and think about data literacy in the modern world sort of holistically. Right? So how might institutions of higher ed and/or industries work together to increase how we're leveraging data, increase data literacy for the future and really help support this data-driven culture that we've talked about throughout the interview?
I think this one goes back to the culture, as you said in the question. This one goes back to the data culture comments that I gave earlier, and for the most part, I view data literacy, at least from an organizational perspective, as a component of data culture.
How do you build literacy? You build that culture. You build those traditions. You build those rituals. You build that shared language and vocabulary, and the literacy will follow. And I'll tell you an example of this:
I've been at organizations who were maybe, you could describe organizations who are emerging in their data literacy and as a data scientist, and I will get asked to attend a meeting. The folks, the others in the meeting will say, hey, we really want the data science perspective, and I'll say, OK, great, I'll be there. Make sure you send me the data ahead of time, so I can review the data.
And in the emerging data organization, this is a bit of a composite experience, but in the emerging data organization, I'll get something that resembles data, but isn't quite data, at least not to my mind. So what's missing? The shared language and vocabulary. Person A asks for data and Person B provides something that Person B believes is data, but in the mind of Person A, isn't actually data.
So that's a literacy problem and a cultural problem, and the solution is for both Person A and Person B to brush up on what is data and then build a shared understanding of what is data.
And I think some folks might disagree with me on this, at least to the extent that I say that if you work on building the culture, the data literacy will follow. And even I myself vacillate from time to time. I think to myself, this is a real chicken and egg problem, which comes first? Is it going to be the literacy, or is it going to be the culture? In other words, if an organization hires a bunch of data literate folk, really highly data literate folk, will they automatically have a data culture? The answer is probably not.
The other side of the coin is if you spend a lot of time building data culture without building that individual data literacy, you may also run into problems. And so the truth is, somewhere in the middle, they go hand in hand. One promotes the other. They both promote each other.
So for everything I've said this entire podcast, my advice for anyone looking to build literacy or culture, specifically as it relates to data, is to have conversations, and the conversations are all centered around the rudiments. They don't have to be very complex, and by rudiments I mean you're going to ask and answer as a group very basic questions. What is data? What is an analysis? How is an analysis different from a data visualization? Is analysis different from research? Does that matter for you and your organization? What are the inputs? What are the outputs, et cetera?
There are no shortcuts. You just have to have the conversations.