I believe there is great upside potential in the TFUAP proposal overall and one feature strikes me as particularly exciting. The proposal to establish “Probability, Statistics, and Machine Learning” as a required area of study in the Science, Math, and Computing (SMC) core could have profound benefits if implemented successfully.
In Mechanical Engineering, empirical evidence informs nearly everything we do – from early-stage design decisions to the analytical work that ultimately brings projects to fruition. My own work is in design, where the role of evidence is especially central. As Herbert Simon famously wrote, “Everyone designs who devises courses of action aimed at changing existing situations into preferred ones.” But how does an engineer characterize the “existing situation”? I would advise any student asking that question to seek and analyze evidence to inform their judgment. How does an engineer evaluate competing courses of action? Again, empirical evidence helps us make decisions that are informed by data and respectful of the uncertainties inherent in real-world situations. And how do we persuade others that a proposal can truly change an existing situation into a preferred one? In many cases, that argument rests on evidence as well. It appears to me that probability and statistics are essential throughout the engineering design process.
The current situation for MIT Mechanical Engineering students presents a challenge. It is possible to earn an undergraduate degree through Course 2, 2A, or 2-OE without taking a subject that establishes foundations for reasoning with data – how it is gathered, how it is analyzed, and how uncertainty should shape our conclusions. Because of this, I have personally observed many situations in which students struggle. A discussion between the student and faculty member begins with a dataset and patterns they both see in a visualization. The student has applied a statistical method they encountered in a lecture in a course without prerequisites in probability or statistics. As the student and faculty member begin to examine the assumptions underlying the proposed analysis, it becomes apparent that the analysis is unlikely to be valid. It’s late in the term and it will be difficult to do all the rework necessary. It’s understandable for students to feel frustrated; they were never taught the conceptual foundations for reasoning under uncertainty. Interactions like these would be far more productive if we shared a common vocabulary and conceptual framework.
Looking forward, the stakes are only increasing. Large datasets and unprecedented computational resources create enormous opportunities for productivity and insight. But without foundations in probability and statistics, these same tools may simply accelerate misleading reasoning and poor decisions.
I strongly support the proposal to establish “Probability, Statistics, and Machine Learning” as an integrated element of the GIRs. I am particularly excited about the prospect of building upon that foundation within our Mechanical Engineering degree programs. Preparing students to reason carefully with data – and to understand the uncertainty that accompanies it – may be one of the most important steps we can take to prepare them for the future.
