Office Hours
Should you get a Master’s Degree in Data Science? A faculty perspective.
You’ve seen plenty of student’s takes. Here is what I think as a faculty member in Data Science.
I’ve seen quite a few posts on Medium (and other platforms) giving a perspective on how to gain the necessary skills to be successful in Data Science. The typical contrast is a university degree vs. a bespoke collection of online credentials (via MOOCs, boot camps, etc.). As a faculty member at Vanderbilt’s Data Science Institute, I think I have a perspective on this decision that is less often heard online.
My background, in brief: In addition to three university degrees, I’ve also taken several online courses and have several credentials from a popular online coding school. I also now have about 20 years of (diverse) work experience, and so I know a fair amount about getting jobs, getting promoted, and being successful. Importantly, I’ve not been a lifetime academic cloistered in an ivory tower — I’ve had deliverables and hired and fired people. If you want more on my background, go here:
Is a Master’s Degree in Data Science worth it for you? Read on and decide. Here is what you’ll get to do in a degree program that makes it worth it:
Develop amazing relationships.
The data science community is large and growing. It happens when you’ve been deemed the “sexiest job of the 21st century.” How can you find your place in the crowd?
Master’s programs work especially well here. You will become a part of a curated cohort of like-minded peers. As you go to class, work on group projects and such, you will begin to develop a valuable network of friends and professional contacts. These peers will help either in the future when they can help you get a contact for a new job, or next year when you are both in new jobs and you need help solving a problem you’re stuck on.
These relationships do not end at graduation, nor are they limited to your contemporaneous classmates.
These relationships do not end at graduation, nor are they limited to your current peers in school. Upon graduation, you will becomes a part of a broader community of alumni. Granted, this is presently somewhat limited as data science programs are still young — there are not many alumni. But this is a benefit that grows with time. I have seen this at work at the Owen Graduate School of Management. The alumni base is active and engaged with the school and ready and willing to help current students and new grads navigate the business world. Step 1 in job hunting is connecting with willing and well-placed alumni.
Even better, you may not realize this now, but there will likely be graduates from your program 10 years from now (or more) that are likely to be great contacts for you down the road, either because you need to hire your next great employee or they are doing something amazing at another firm and you see a possible partnership. You have an automatic bridge to reach out to them through you common school connection.
Learn from your talented classmates.
Don’t underestimate the importance of learning from peers. This is one of the most important parts of education, really. There is significant research that shows how effective peer learning can be. Here is the basic explanation: a peer who has just recently understood a concept is much closer to her peer who is still confused. Thus, this person is more likely to use specific words, concepts, or metaphors that aid understanding. Teachers, who are experts, can sometimes be at a loss to explain something in a new and fresh way that will work for some students.
Further, teaching someone else is an excellent way to build expertise in a topic. Hearing and understanding is very basic. Applying that knowledge to a project or problem is a deeper level of understanding. Teaching others to do the same takes you even further.
Working in a community of learners/scholars produces much deeper understanding than studying on your own.
Also, the faculty are amazing. I don’t mean this as self-aggrandizement — I’m talking about my peers. They are amazing, talented people who are fully engaged with students and care a great deal about student success. So peer learning is awesome, but you are never left on your own.
Stand out in the crowd.
One of the most relevant reasons to get a university degree is that you really stand out. Online courses often have thousands of participants. Web-based platforms to learn coding likely have hundreds of thousands of learners. This is great! I’m thrilled that there are so many who want to learn more and I fully support these platforms (a topic for another post). As I noted above, I’ve used these platforms myself and benefited from them.
However, once you finish, you are among thousands that have just done the same thing. In a year or more, you are one among perhaps a hundred thousand. How will you stand out to a recruiter or a hiring manager?
Graduate programs, by necessity, have screening built in. It is called admissions. While this process feels arduous when you are entering it, once you are admitted, it is rather helpful. This process is what helps you stand out as a part of a school and also helps generate an amazing peer group of talented classmates.
Practice working on a team.
I often say this in my own courses at Vanderbilt, but data science is a team sport. One of the things you will definitely experience in a degree program is working in teams. This kind of experience is vital because virtually all work you’ll need to do professionally will be as part of a team. You will have your own individual work, of course, but unless you are a one-person company, you’ll be working with others eventually. It is better to practice that sooner rather than later.
How do you divide the work? What do you do if someone is not doing their part? How do you avoid duplicating effort? How do you effectively communicate specifically what needs to be done? How do you do all of that with time pressure and a deadline?
How do you work on a team?
These are all things you will experience in a degree program first hand. These are also things you will experience in the real world of work and deliverables and deadlines. It is good to start now.
Prioritize deadlines and work tasks.
Yes, deadlines are a thing. To be clear, I don’t just mean making sure you finish your work by a certain deadline. I also mean managing multiple project or problems and multiple deadlines all at the same time. I mean managing your own individual deadlines along with your team’s deadlines in a different class.
In the work world, you can’t always just power through everything and get it all done. Sometimes, you need to ‘manage upward’ with your supervisor to let them know that some due dates for deliverables need to shift. How do you have that conversation? What is the right way to discuss this? How do you prioritize with your supervisor?
I regularly talk with students who are managing a lot — not just classes but interviews (job hunting takes a lot of time and happens concurrently with your classes), case competitions, and other commitments. I do my best to be flexible. Something I can deal with is a student who emails me asking about a project deadline in a week that corresponds to a midterm in another class. I often do not know about conflicts like this. If my project was due on Friday, I can often move the deadline to Sunday with very little impact on my schedule, and students can now focus on their Friday exam, then pivot to the project conclusion on the weekend.
Practicing this kind of advanced planning and time/workload management is a part of being a graduate student. It is a vital skill in the world of work. You get these kinds of situations when you are studying on your own.
Develop a deeper understanding of…everything.
Rigor is probably my biggest concern with respect to automated learning. Let me define that: by “automated” learning, I specifically mean self-paced learning where there is no personalized interaction or feedback. This is primarily found in online MOOCs with thousands of students and the automated platforms that teach coding through self-paced videos and simple exercises.
I have found it impossible to combine rigor with automated learning, at least with current tools and technologies. I have many automated assignments in my classes — readings, videos, and auto-graded quizzes and feedback. They are an important part of my class…for basic introductory concepts. My students do these as prep work. The in-depth learning comes in class via discussion, group problem solving, and the like.
Rigor comes in (at least) two forms. First, it is simply dealing with harder problems. You cannot automate harder problems because too many students will get stuck and automated feedback is impossible unless you know the error. Right now, it takes a human to see what the student is doing wrong and customize the help to move them closer to the answer without giving it away. Automated platforms also always err on the side of ‘easy’ because nothing is more frustrating for users than not being able to solve a problem and not having anyone to ask for help.
Rigor is not just harder problems. It is open-ended problems.
The more important form of rigor is open-ended problems and projects. For example, as a final project in my Survey of Data Science Applications class, my students are writing their own case study on an application of data science. This assignment does a lot. First, it tests their understand of specific uses of data science in real life, forcing them to research how firms or organizations are using data science. Second, it puts them into the role of “teacher” because they have to identify learning objectives and class discussion questions. This pushes them to a higher level of understanding as I noted above. Third, it also tests their communication and writing skills because they have to bring all of their ideas together in a concise document. There is no way this kind of assignment would work in any automated learning platform. It pushes my students in ways that those programs simply cannot do, and thus engenders a deeper mastery of those topics.
Work on unique, interesting, and important projects
If you read for about one minute about “how to be a data scientist” you will come upon advice that you should work on projects.
Shortly after that advice, you’ll find more advice that you need to stand out from the crowd, so your MOOC final project (also done by 10,000 other students) won’t work. Your slight variation on a Kaggle competition isn’t likely to draw attention either. And you definitely shouldn’t use any data set you’ve ever seen in a tutorial online.
Most Master’s programs will include working on interesting projects in conjunction with faculty or local businesses. Often, they already have interesting data and an array of real questions or problems they are interested in solving.
A Master’s degree should get you working on real projects with real data and real obstacles.
But you also have real support and the resources you need to succeed. This kind of experience is exactly what you need.
These are the types of projects that help you stand out. These are the types of projects that help you build your skills as you deal with messy data, badly named database columns, and the disappointment of realizing that what you thought was your key variable is missing in 75% of your data because it wasn’t a required field.
Engage in meaningful conversations about ethics and bias
Most programs have a required ethics course. You can (and should!) read about the ethical issues in data science. There are quite a few and I won’t go into them here (e.g., racial bias, data privacy, etc.). Engaging in difficult conversations about ethics is best done in person and repeatedly through time. They are also best had within a broader context, not in isolation.
The ethical use of data and ethical application of algorithms (for example) are the responsibility of everyone in the field. We can’t push ethics off to the side as the specialty of a few.
Ethics is something that is easily omitted from a self-curated set of certificates. It is also virtually impossible to do in an automated learning environment. My peers at Owen that teach ethics were particularly distraught at the move to online with the pandemic because of the loss of these conversations.
Much of ethics is understanding how your own seemingly innocuous choices can badly impact others. Unintended consequences. Without others to help you see your own blind spots or tell their stories, it is hard to grow your ethical imagination.
Finally: Let’s talk about cost.
The cost of the degree, of course, is the main issue here. However, rather than looking only at the cost, you should ask yourself if it is a good value. The costs are pretty tangible in the form of tuition. I’ve tried to describe the benefits, which are perhaps more abstract.
So, to finish up, I’ll spend some time talking about why these degrees are so expensive. It isn’t because universities want to gouge you —they are non-profits after all. As far as I can tell (this is above my pay grade), most university programs run at cost; they just want to break even.
Why does it cost so much?
First, good teaching is costly. Why are you interested in data science? I’d be willing to bet that at least part of that interest relates to the pay scales. Given what an experienced data scientist can make in industry, what do you think it costs a school to pay one to teach full time? Don’t you think that data science instructors can get pretty nice jobs, especially given the current demand for those skills? Most professors make less as a professor than they could make in industry, but the lifestyle (more flexibility, more autonomy) is a non-monetary benefit. Personally, I also really enjoy the variety, the relationships and the freshness that comes with each new cohort of students. And nothing makes me happier than hearing about alumni doing amazing things.
This is broadly true. The more a profession pays, the more schools have to pay the teachers, even if most of us are willing to work for less than industry salaries. But…it is just a bit less, not a lot less. So, business schools, law schools, medical schools, etc. all have higher tuition because it costs more to employ people with those skills. The faculty have lucrative outside options, so the salaries need to keep up to keep your best faculty.
Good teaching is costly. Good teaching does not scale.
Second, good teaching simply does not scale. I’m a big believer in Ed Tech and love many of the new technology tools, but I hope I’ve convinced you that a human teacher interacting with students is still the best way to develop depth of understanding in complex topics among students. Further, the relationships you build with fellow students are vital not only to your understanding but your long-term career development. I have not yet seen a technology that can replicate this experience in a class of thousands.
I’ve seen this myself because I have written and designed automated tools for my classes. I use them to introduce concepts and provide accountability around class preparation. But automated tools do not work well for anything rigorous or anything open-ended, which are the most important and interesting questions. Automated tools are not good for developing depth of understanding. And without automation, you can’t get scale.
Is it for you?
I hope so. I’m a big believer in education in general, and higher education in particular. That is why I do what I do.
Higher ed is not for everyone, but if you are looking to break into Data Science and willing and able to make the necessary investment of time and money, I believe it is worth it.
