RETHINKING THE DATA SCIENCE DEGREE
The U.A. system retools to teach real-world innovation
Karl D. Schubert, Ph.D
Research Professor and Director of Innovation and Data Science Initiatives
College of Engineering and the Sam M. Walton College of Business
University of Arkansas
THIS STORY BEGINS about a decade ago, when I was invited to participate on the College of Engineering Dean’s Advisory Council, established in 1989, whose members are engineering alumni and national industry leaders who are committed to the College’s pursuit of excellence in research activities, scholarship, and academic programs. The Advisory Council’s goal is to ensure personal and professional growth for future generations of engineering leaders.
As it turned out, we on the Advisory Council discovered something disturbing: It appeared that the engineering students, over the years, had become less and less able to solve problems that they hadn’t already seen. Also, they weren’t very good at working on teams. I was asked to head a sub-committee to look harder at these issues, and what we basically concluded was that the Engineering College was training the creativity and innovation out of the students.
How was the College doing that? Well, by giving these students a toolkit and saying this is the process you use to solve problems—and more or less sticking to that method without ever giving them open-ended problems or problems involving teams. As a consequence, these engineering students had become used to having everything handed to them, which is a perilous way to prepare for real life.
The first thing we did was to discuss our findings with Arkansas companies that were potential employers of our engineering grads. “We need to fix that,” they all agreed. But we soon found that this was only the spear tip of the issue. One day Engineering dean John English and College of Business dean Matt Waller were talking, and Dean Waller said, “We’ve got the same problem in the business college.”
So the two deans retained me to analyze the situation further and then to develop a way to reintroduce innovation back into the program. In the bigger picture, the idea was to create an innovation concentration in which students from both the College of Engineering and the College of Business could take courses together that would include real-world problems and a team-based, student-led, open-ended way of approaching these problems, through innovation and creativity.
This pilot program, which I developed three years ago with a courageous instructor, Mrs. Leslie Massey, was called the Honor’s Innovation Experience. I started with honors students because they can be more flexible—plus they’re oftentimes among our best students. I knew that if we told them the class was an experiment, they would respond to that—and they did, putting in extra hours to help make the experiment a success.
At first it was just the engineering college—a two-semester class with 20 students. First semester we talked to the students about the innovation process and the innovation ecosystem. They ultimately formed teams and picked a project to work on. The students basically had to do the project themselves, to the point of doing a market analysis, a definition of the problem, customer interviews, minimal viable product, cost estimate, some scaling information, and building proof of concept (if it was possible to do so for that project). We had a successful first year.
The second year we doubled the class to 40 students, and this time we made a major change in our process. In the first year we had found the mentors first, and the teams were then given a project by the mentors—all except for one team. That particular mentor said, “No. I’m not going to do it that way. What I want to do is say, ‘Here are the areas I work in. If you pick a project in my area, then I can mentor you.’” That mentor’s team did an outstanding job, while the other ones were just kind of okay. So, in year two, we turned it around. The students came up with the project they wanted to do, and then we found mentors to match the projects.
In this third year that we just finished, we added business students to the mix. We had 30 engineers and 20 business students in the program, the idea being that we would form teams of three plus two and do the same kinds of ambitious projects. This pilot program has gone very well, and this coming year we are increasing the number of business students to 30 for a total of 60 students in the program. And the overall First-Year Engineering Program will also be integrating several of our modules into their classes for the benefit of all College of Engineering students.
With that as background, in our discussions on innovation with local, regional, and State-based companies the topic of data science kept coming up. Many of our companies—current and potential future employers for our students—depend on data for their success and growth. As a result, the deans of three of the UA system’s colleges—J. William Fulbright College of Arts & Sciences, under Dean Todd Shields; Sam M. Walton College of Business under Dean Matt Waller; and Dean John English’s College of Engineering—got together and decided to put together an exploratory team to understand this need and, if appropriate, to develop a curriculum to meet it. I was asked to chair that committee, and 15 months later, we are in the faculty review and approval process (curriculum is faculty led). We’re hopeful that our new curriculum will be approved as an official B.S. Data Science Major as a collaboration of the three colleges. Our target date is Fall 2020, with an opportunity to start earlier if funding and support are available.
I COME FROM the business world—from Dell, IBM, Honeywell, mid-sized companies and start-ups. I have decades of experience working with data and systems, and I’ve been a senior executive at companies. I’ve worked on the technical side, and I’ve worked on the business side. I’ve been a general manager. I’ve managed $80-million divisions of companies. Data makes the world go round and, to me, without software and without data to do something with, a server is just a space heater or a doorstop.
That’s why today I have a tagline that appears on all my business communications: “Connecting technology to business value.” That insight was sparked by a conversation I once had about the difference between engineering and the sciences. “In the sciences,” I said, “the question is whether you can do something or not. In engineering it’s, ‘Can you do something to make life better for people, and can it have a business value to it?’” That’s how that tagline came about, and I’ve used it ever since I began working. I think it’s especially important for these students to understand that data and their talents are a means to an end, as opposed to just an end in itself.
In fact, this whole years-long process began with intense discussions between members of our Data Sciences Curriculum Development Committee and senior business executives throughout Arkansas. The business people were very specific in what they wanted a new Data Sciences curriculum to instill in their future tech employees. It really came down to six outcomes, of pretty much equal importance.
First was “Use of technologies for solving real-life data problems.” In other words, you can’t just rest on your tech knowledge per se. In a regular computer science curriculum, you could be taught how to create a file system, how to create a database, how to create an operating system, how to create a compiler, how to write things from first principles. But in our program, you’re going to use those things to get to the data. What’s important is making sure that the data’s right, that it’s clean, that it’s valid, and that you can then do things with the data. The example I use is, in the past you learned how to build a car. Now what we’re going to do is teach you to drive.
The second desired outcome was “Ability to develop models and draw conclusions.” It isn’t enough just to be able to analyze the data. Our students will be able to create abstractions at a systems level, develop models from that, and then use those to draw conclusions from them.
Number three is “Critical thinking and problem-solving.” I use the example of an old-fashioned slide rule, an unfamiliar relic to most of today’s students. A slide rule helped a user with the digits, but not with the decimal point. So, once you ran through your engineering problem, you had to look at it and figure out the power of ten in the answer. You had to figure out whether or not that made sense—and if it didn’t, then you might have something else wrong, too. In critical thinking, you’ve got to be able to look at things and say, “Is this reasonable? Is this rational? Does this make sense?” And also, “What are the societal and ethical impacts of what we’re doing?” All our students will be required to take social science courses, because they need to understand the social and ethical implications of their work.
Number four is “Interpreting and implicating.” You can’t just report the data—you’ve got to be able to draw conclusions and even project from it and be able to give that to senior decision-makers. Otherwise, what’s the value of the data?
Number five is “Communication skills.” Engineers and scientists have a reputation for lacking such skills, especially outside their own discipline. Our data science students will be able to communicate their knowledge of a project to someone who perhaps speaks a different “language”—a CEO, say, or a marketing director, or an engineer, or the public.
That leads us to number six, “Teamwork and knowledge transfer.” When our students graduate, they’ll already be used to working in interdisciplinary teams. At the university, I have to say “multi-college and interdisciplinary” because engineers believe that working with other engineers is interdisciplinary, which is missing the point because to non-engineers that does not appear to be interdisciplinary. But by having worked in these multi-college teams, they’ll start out in the business world understanding that everyone brings something to the table. In the Honors Innovation Experience pilot program, for example, I learned that there was a certain practicality that came from having the business students in the class with the engineers—the business students tended to ask really good questions that caused the engineers to think about cost, and how everything fit together. And, the business students benefited from the practicality and doability brought by the engineering students.
The important thing for me, personally, is that this ability to work with people from outside their field will give our students an edge career-wise over their peers from other schools. They’ll already find it normal to seek out people in other departments and disciplines to work together in solving problems.
TO MAKE CERTAIN that all our Data Science students are steeped in this kind of thinking, we designed this new multi-college interdisciplinary curriculum as a “hub and spoke” system, in which there will be a core set of classes that all the students take, and then a series of concentrations that are 20 to 21 hours of the degree that are specific to their chosen field.
The B.S. Data Science Core includes 35 hours of general education in the areas of math, science, humanities, fine arts, and social sciences. The Core also includes three key elements for Data Science:
- Computing and Programming Foundation: Data Science lingua franca (R, Python); object-oriented programming (JAVA); programming algorithms and paradigms; data structures and databases; data processing; and cloud computing.
- Statistics and Probability Foundation: Probability and statistics; multivariable math, including linear algebra; statistical methods for Data Science; decision making; machine learning; and optimization.
- Multidisciplinary Environment: Technical composition; the role of data science in today’s world; micro- and macroeconomics; business foundations; data visualization and communications; social issues in Data Science; and a mandatory two-semester multi-college interdisciplinary practicum.
Beyond the Core are 10 concentrations: Accounting Analytics; Bioinformatics; Biomedical and Healthcare Informatics; Business Data Analytics; Computational Analytics; Data Science Analytics; Geospatial Data Analytics; Operations Analytics; Social Data Analytics; and Supply Chain Analytics.
Finally, every student will be part of an interdisciplinary team tackling a real-life capstone project. The plan is to get companies who would like to participate with us to give us real problems with real data for the students to work on. For example, I recently mentored an industrial engineering capstone project team whose problem, from the industry partner, was, We need to improve our profitability by one percent. We don’t know how to do it, and we have millions of data records on our sales, but we need your help to figure out how to do that. It’s a very open-ended problem. It’s a high-level definition, and it requires the students to be able to talk to people outside their domain and to research, identify, and use approaches completely new to them.
At first, they were scared to death. It was like, “Oh, my God!” I tried to help them learn how to think about it. “This is going to be the kind of problem you’re going to get when you go to work,” I said. At that time, they were eight or so months from graduating. “Then think secondly that it’s so broad that you actually can define what the solution is.” It’s also an open-ended problem, which brings its own challenges.
The team set up weekly calls with their interface at the company and met regularly as a team, so they were in good communication. I met with them every week to give them advice and counsel and they sought out subject matter experts from faculty and graduate students in specific areas of need.
In early April, after they had completed the project, I said, “Looking back at it now, how do you feel about it?”
“It actually wasn’t that bad,” they said. “You were right.” Now when they go out to get their first real problem to work on a team and also to work on a less than perfectly defined problem, they’re not going to freak out because they know there are paths to dealing with it. They just have to find those paths.
IN ORDER FOR the Data Science program to become official, we have to get approval from the individual departments, then the curriculum committees of the departments, then the colleges, then the university curriculum committee, then the faculty senate, and finally the Arkansas Board of Education and the UA Board of Trustees. You may wonder how such a thing can ever happen—all those different people, all those different priorities and interests!
But I would be remiss if I didn’t tell you that we’ve been moving through the approval process for about a year now and it’s going great. I wish I had space here to praise every single person on both the Data Science Advisory Committee and the Data Science Curriculum Development Committee.
The Advisory Committee is comprised of engaged and proactive leaders from our state’s business, government, and academic communities, and these people have done an amazing job in pointing the way to the future for us. Then the Curriculum Development Committee, for which the deans of the three colleges recruited people from their teams who are collegial but are also the right people in their respective colleges and departments, took our interviews, their knowledge, with benchmarking of other universities, and the Advisory Committee’s blueprint and turned it into reality.
None of this, I must say, could’ve happened had not the deans of the UA System’s three colleges worked so extraordinarily well together. This is unique—it sure wasn’t this way when I went through school here. But when we kicked off this process, the three deans came to the meeting and spoke eloquently, and from the heart, about how important this curriculum change will be to our employers in the state and the region; to the university and its continuing reputation for excellence; and certainly to our students, who will graduate ready to make major contributions to their new employers.
Thanks to these forward-looking leaders, we’re now on the brink of launching a ground-breaking program that will position our Data Science students to be sought after by local, state, and regional companies. Only two other universities in the entire country—Ohio State and North Carolina State—have done anything remotely this ambitious, and we visited both schools and learned from their experience. This is the future, and it is nearly here.