As the data revolution transforms industry and society, engineering schools are rethinking the basics.
By Mary Lord
The world is awash with information. Each second, a vast array of increasingly intelligent devices pours binary tidbits into the metaverse, conveying details from the mundane (email, texts, Facebook posts) to the vital (medical scans, tornado alerts, aircraft malfunctions). This decade-long deluge, dubbed “the golden era of data” by Harvard Business Review, has transformed society and made data skills essential at every level in virtually every organization.
That includes higher education, where the data revolution is driving discoveries in areas from archeology to vaccine development. The impact and opportunities have been particularly profound in the STEM arena, as evidenced by the gleaming, multidisciplinary computer, data science, and engineering complexes springing up on US campuses. Industry’s surging need for graduates who can effectively and ethically winnow insightful kernels from the chaff of enormous data sets propelled the National Academies of Sciences, Engineering, and Medicine to recommend in 2018 that academic institutions embrace data science as a “vital new field” and create majors, minors, and other undergraduate pathways to foster “data acumen.”
Schools have responded with a flurry of novel offerings. Some even overhauled their entire programs. In 2018, for example, Rensselaer Polytechnic Institute became the nation’s first university to require “data dexterity” for all students as part of its updated core curriculum. The University of Utah features an undergraduate certificate in data fluency that emphasizes basic “data wrangling” and ethics, while Purdue University’s living-learning community Data Mine promotes data science across majors and schools.
Despite their crowded curricula, engineering schools also are joining the Big Data movement. In the process, many are generating lessons to guide efforts on other campuses. The University of Michigan was an early proponent, launching a joint data science major between the electrical engineering, computer science, and statistics departments in 2015. Other engineering programs have followed suit, forging multi-college data science bachelor’s programs like the one the University of California–Davis will roll out this fall. Data science minors within engineering likewise have proliferated, as have pathways within departments, such as the data science option within the University of Washington’s industrial and systems engineering degree program.
Fresh Foundations
Few programs have retooled their undergraduate curricula for the digital economy as dramatically as Boston University’s College of Engineering. Four years ago, the school began embedding data science concepts in required courses for all majors, putting statistics, probability, and basic machine-learning concepts on par with calculus among fundamental competencies. An expanded core sequence now is poised to debut, with upper-level classes in data analytics, artificial intelligence, and other subjects. The goal: graduate innovators who understand and can apply large-scale computing techniques to global challenges from improving medical care to designing sustainable cities.
“Solving complex problems for society requires intersectional thinking,” explains BU’s engineering dean, Kenneth Lutchen, who spearheaded the initiative as part of the school’s trademark mission to educate the Societal Engineer. Students “still are card-carrying mechanicals” and other majors, he assures, but “every single engineering discipline needs engineers who know something about data science.” Given the profession’s code of ethics, he adds, engineers have “an extra obligation” to “look at the data and say something isn’t right.”
BU’s big-data shift builds on a decade of curricular transformation, such as infusing makerspace and design experiences throughout the undergraduate engineering program. It also responds to the “data science wave” that the dean’s leadership advisory board saw reshaping their businesses. The shared top need of biotech labs and retail giants alike: engineering graduates with exposure to data analysis tools and techniques. The “shock” that further validated the school’s data science commitment came when Lutchen was interviewing potential hires this past year and realized how deeply data science “is impacting every one of our faculty.” Of ten recent candidates, for example, two explicitly worked at the intersection of data science and medicine. The rest (like dozens of previous applicants) used machine learning and artificial intelligence to advance research in such traditional engineering areas as materials design and energy.
Inside the ‘Black Box’
Spurred by this end-user feedback, Lutchen established a committee to examine the undergraduate curriculum and recommend revisions. The core data science sequence that emerged was rolled out college-wide in fall 2018. Initial reforms targeted “low-hanging fruit,” acknowledges Lutchen. All freshmen used to do MATLAB projects, for example. The redesigned introductory course adds Python, a popular data programming language, and emphasizes applications engineers would employ on the job. A required two-credit linear algebra class was remade into a three-credit computational linear algebra course filled with such real-life examples as Google’s PageRank algorithm.
Identifying inefficiencies opened schedules and improved learning. For years, each engineering department taught separate probability and statistics classes that covered the same material. Today, all majors take a common four-credit course—Probability, Statistics, and Data Science for Engineers, or EK381—that not only introduces data science concepts but also provides a solid foundation for upper-level electives in data analytics, machine learning, and other topics that draw heavily on probabilistic reasoning. Result: equally well-prepared students, regardless of discipline, who understand enough about modeling and analyzing complex systems to pursue data-intensive emerging fields from autonomous vehicles to personalized medicine.
Like their predecessors, these reimagined foundational courses emphasize mathematical problem-solving and conceptual thinking. Their purpose, however, is to illuminate how algorithms—which abound online—and other data-crunching tools work rather than train students to develop systems that compute answers. “I focus on teaching them what happens inside the black box,” explains EK381 instructor Bobak Nazer, an associate professor with expertise in information and data sciences in the department of electrical and computer engineering who oversaw the design of the college’s revamped core. That means starting with pencil-and-paper examples and building toward more theory-based, machine-learning applications via visualizations and formulas. “We’re always trying to live at the abstract level,” Nazer underscores, so students learn “how to stitch algorithms into something they can use that’s interesting” as engineers.
Cats, Dogs, and Mammograms
EK381 ignites that interest through engaging image-classification case studies. The electrical and computer engineering case study, for instance, asks students to write MATLAB or Python code to determine if grayscale photos show cats or dogs. Other challenges involve predicting concrete compressive strength and distinguishing benign from malignant tumors on mammogram scans. “Students aren’t diagnosing cancer, just looking at features of tumors to build a prediction,” explains Nazer.
The course uses real data sets, curated so students can’t manually sort through the images yet small enough that one tiny programming mistake won’t tie up their laptops for a week. It also contains “a lot of pencil-and-paper problem-solving and theory—you can’t get away from that—so students see why they are doing this,” says Nazer, who developed EK381’s course materials with David Castañón, a professor of electrical and computer engineering. Much of that math-heavy calculating is done in the flipped classroom, thanks to Khan Academy-style online tutorials Nazer made during the pandemic. (Access the videos and other course materials at http://probstatdata.bu.edu.)
“It’s a very intense class,” attests Compton Bowman, an aerospace engineering major who attended community college and played in a band before pursuing his dream of inventing a helicopter-like space shuttle. Had he faced such an “assault on all fronts” fresh from his meager high school math and science offerings, he “immediately would have left.” While Bowman struggles at times—“you can’t learn data science by ear,” he quips—the course’s “brilliant” instructors and abundant study supports have kept him buoyant. Data science has yielded “revelations,” particularly about assessing safety and risk, says Bowman. “I can’t think of an engineering application where it wouldn’t be useful.”
Comprehensive Support Systems
EK381’s innovative structure offers a model for rigorous, equitable, and inclusive data science integration. Consistency is a hallmark: Each section’s instructors teach the same syllabus using the same textbooks, homework, and exams. As a result, learning outcomes are “similar across majors,” reports Nazer. While a background in computer science or statistics may give some students an early advantage, he adds, a strong foundation in algebra and multivariable calculus helps most.
The curriculum is constructed to instill an intuitive feel for statistics and probability—unfamiliar subjects for many students. Miriam Bounar, a pre-dental junior majoring in biomedical engineering, had never taken probability and anticipated a slog in EK381. Instead, she shaded in rectangles on her first exam and spent class calculating the odds of drawing four aces from a deck of cards. “It doesn’t feel like work and actually is kind of fun,” Bounar marvels. She also now sees the course’s relevance to her other classes. On a recent cell biology exam, for example, she applied what she knew of probability to calculate the total possibilities for a reaction to occur—as she’ll need to in real-life practice.
Nazer deliberately encourages questions and strives to answer them all. “Everything is a teachable moment,” he says. Bounar appreciates his low-stakes approach. “No question is too far-fetched,” she says. Unlike other classes, where “it’s awkward, and a lot of pressure, and you don’t know what to ask,” EK381’s lively discussions and diverse perspectives help her “grasp the materials more.”
Students also can tap multiple sources of assistance. Zoom options make office hours easier and more convenient to attend. Online forums help unscramble pre-exam muddles, while YouTube videos showing the professor’s step-by-step calculations let students review and prepare for in-class problem-solving.
Dogancan Kuyel is a fan. The biomedical engineering junior finds it “difficult to process” if a lecturer jumps from topic to topic. With videos, he can pause and replay until he nails how to do the proof. “It’s absolutely beautiful!” An artist with a passion for synthetic biology, Kuyel says the course’s blend of theory and everyday examples “really sparks curiosity.” It also “is not just some testing facility—and with engineering it’s easy to feel like that.”
For many students, one of the most useful data science support systems involves the use of undergraduate teaching assistants, a practice Nazer encountered and borrowed from his undergraduate alma mater, Rice University. As Bounar explains, these near-peer TAs “know what we’re thinking, they got confused in the same places, and they’re going through the college experience with us,” which fosters rapport and effectiveness. They also suggest course improvements, such as providing more information about Python and other popular data-science programming languages.
Undergraduate TAs often draw on what helped them succeed. To demystify conditional variables, for example, EK381 TA Julia Roy, a biomedical engineering senior, “works through difficult concepts together [with students] rather than just explaining everything at once.” As a TA in the first-year Introduction to Programming for Engineers class, Kuyel typically proffers personal time management tips, such as how he would tackle a course, plan a schedule, or devote hours to studying a few pages rather than skimming the textbook. Ultimately, says Kuyel, “I just want to see these students learn—to say engineering is cool, not hard.”
Expansion Plans
BU’s big bet on Big Data to inspire learning, expand career opportunities, and spur human-centered innovation extends beyond the engineering curriculum. A stunning, 19-story Center for Computing and Data Sciences now towers over the central campus. Slated to open this year, it puts the Rafik B. Hariri Institute for Computing and Computational Science & Engineering under the same roof with a new data science major and the mathematics, statistics, and computer science programs. The engineering school’s data initiative also is poised to break fresh ground. A new concentration in machine learning has been approved, which can be added to any traditional engineering major. Plans call for identifying and leveraging places in the curriculum where data science and machine learning can augment subjects such as energy, electrical engineering, electronics design, neuroscience, and medicine.
Whether other engineering schools will mimic BU’s core makeover remains to be seen. But EK381 course evaluations suggest students appreciate the changes. Before his data science immersion, Bowman “didn’t have that certainty, that ability to see myself working as an engineer.” He’s now sure of his path and understands “exactly what I need to work on” academically to follow it. His probability of success: “Not zero.”
Mary Lord is Prism’s deputy editor.
Design by Nicola Nittoli