Zach del Rosario '14, is teaching Data Science and ModSim this fall.
Uncertainty permeates all facets of our reality. We incorporate assumptions of uncertainty as we go about our daily lives, manifesting, for example, in interpreting the weather forecast with a healthy dose of skepticism or beginning a commute ten minutes earlier than your navigation app would lead you to believe is necessary. The idea that uncertainty abounds in our lives is hardly controversial, especially to the reader in the fall of 2020 who is bombarded daily with new COVID-19 case statistics accompanied bycompeting analyses, simultaneous claims of excessive and under-testing for the virus, and more.
The COVID-19 pandemic provides an excellent opportunity for both statistical and semantic reasoning about data to not just support conclusions with statements of fact but arguments for which conclusions can be drawn confidently and which cannot. Zach del Rosario, the instructor for this fall’s Data Science course at Olin, shared that one of the class’s projects will entail this kind of analysis with coronavirus-related datasets. “There are a whole host of borderline unquantifiable ways in which the [COVID-19] data are wrong,” argued Zach, citing the varying rate of testing over time as just one of the confounding variables distancing the reported data from ground truth. “It is sometimes a deeply unsatisfying… conversation to have because you are not going to come to some exact solution at the end of the whole thing… It’s instead, ‘Well, this is our best impression of what’s going on right now based on all of these caveats. What can we confidently conclude?’”
Zach graduated from Olin in 2014 with a mechanical engineering degree and began pursuing a PhD in aerospace engineering at Stanford. During Zach’s internship at Northrup Grumman in 2017 where hewas searching for a topic for his thesis, he had an ‘aha moment.’ “I was going around the company basically interviewing people using a lot of the playbook that we learned from UOCD [Olin’s User-Oriented Collaborative Design course] to… figure out, ‘What do real engineers do when it comes to aircraft design?’” Zach’s epiphany was his realization that the material property values he was trained to treat as deterministic were, in fact, not.
Mechanical engineers rely on material property values in a variety of scenarios, especially when creatingsafety margins in their designs. To account for variability in the physical composition of a material used in manufacturing, it is convention to treat it as homogenous and design with ample margins of safety at every level. While this has proved to be a successful approach in many scenarios, Zach has foundthrough his research that this process, when applied to composite materials, can yield structures thatare not as strong as intended. Juxtaposing this finding with the aerospace industry making increasing use of composites leads to an alarming conclusion: the production of aircraft that are unsafe due toinsufficient designs is an increasing problem.
More than just articulating this problem, Zach developed alternatives to traditional structural design criteria that make use of rigorous statistics and have provable safety properties. Tying back to his realization at his Northrup Grumman internship of the indeterministic nature of material properties,Zach described, “The thing that I’m really pushing on in my research is trying to help engineers get beyond a purely deterministic view of their problems.”
Zach also recognizes the importance of effective communication in realizing the full impact that researchfindings can have – something, he thinks, is a point for improvement among scientists. “You probably couldn’t pay most people to sit in front of a scientific talk,” Zach explained. “But people will gladly go and pay $19.50… [for] a movie. And if we strip both of those things down… you’re really putting information in front of people… [but] the value flow is perceived as being very different here.” Pointing to Houston, We Have a Narrative: Why Science Needs a Story by Randy Olson as a point of inspirationfor his emphasis on communication, Zach described that one of the techniques for effective communication is the “and, but, therefore” framework which organizes information into the exposition, conflict, and resolution elements respectively. The resolution element, Zach claimed, is what scientists often leave out to the detriment of making sense of contradictory evidence and leaving the reader with a punchline.
These communication skills are something that Zach is incorporating into the Data Science course: “With data science, it is… [the use of] statistics and computer science to try to make sense of data and we’d really like to use data science in order to support decision making. But, if I give you a random pile of facts from my data science, you’re not going to be able to make any sense of it. Therefore, I teach students narrative storytelling as a way to try to help them tell their data story.”
Accompanying communication in the – as Zach describes – “toolkit to reason about uncertainty” is not just statistical literacy, but also the skills to ask the right questions. When making a point about asking questions in data science, Zach cited a quote from the statistician Jon Turkey: “Far better an approximate answer to the right question, which is often vague, than an exact answer to the wrong question, which can always be made precise.” Indeed, the very approach to addressing uncertainty is likely to incorporate uncertainty itself.
Beyond this semester’s Data Science course, Zach hopes to make a broader impact in the engineering community with regard to the treatment of uncertainty. “Let’s take that general problem [a lack of appreciation for uncertainty] and look at it across a lot of different disciplines,” Zach described, following an emphasis that he is aiming for broader implications than composite airplane design. “Let’s figure out how to fix that from the pedagogy side. That is the kind of problem I am looking to work on while I’m at Olin.” If you are interested in joining Zach in this endeavor, he is actively looking for student assistance; you can find out more at https://www.zdelrosario.com/students/.