This academic year, I am a pedagogy fellow (PF) at the Harvard School of Public Health (HSPH). Instead of being a TF/TA for the Biostatistics Department (which typically involves grading, teaching labs and holding office hours for students), I work on developing curriculum and encouraging evidence-based teaching practices, both within my department and for the school overall. For HSPH-wide projects, I collaborate with PFs from other departments. In this post, I'll focus on insights inspired by my department-specific project as well as by a course on teaching that all the PFs took last semester.
For context: in spring 2020, I really enjoyed taking a course at CU on statistical collaboration. (The Professor, Eric Vance, used the term collaboration as opposed to consulting to emphasize the importance of consistent, two-way communication between statisticians and domain experts.) When I heard that my current department was looking for a PF to redesign our PhD-level statistical consulting course, I applied immediately.
Beyond digging into statistical consulting, being a PF has been a great opportunity to think deeply about whether I want teaching to be a core part of my future career. The upshot on that front is that I'm still making up my mind. However, many of the ideas I've been synthesizing are broadly applicable on both individual and systemic levels, sufficiently so that they merit a post. The three main topics I will discuss are (a) evidence-based tools in education, (b) key skills common to teaching and consulting / collaboration, and (c) broad musings on modern education.
Evidence-Based Tools in Education
At the start of Teaching 100 (the course the PFs took last semester), we read the book Make it Stick: The Science of Successful Learning. We then proceeded to have eight weeks of discussions, led by instructor Tari Tan, about this material and more. During the class and in the months since, a few concepts in particular have stuck out to me.
The number one takeaway from Make it Stick is that retrieval practice is critical for long-term learning. Retrieval practice is honestly testing your knowledge (as opposed to re-reading material), ideally waiting long enough between practices that you start to forget. Learning both concepts and skills this way is intentionally challenging, but has been shown over and over again to yield better educational outcomes than traditional studying. I was introduced to retrieval practice through studying music (my teacher would have me do practice performances where I couldn't look at the sheet music), but looking back on my academic experience, I think retrieval practice is still under-utilized in most traditional classroom settings.
A related concept is one that many people nowadays are familiar with: the power of cultivating a growth mindset. Despite the proliferation of this idea, STEM education (and math in particular) struggles with deeply-rooted notions that some people are destined to flourish and others are doomed to flail. My personal experience in math classes from middle school through grad school is that it can be really hard to prevent unsuccessful attempts from harming your perception of your overall competence (see "Potent Challenges" in a past blog post). Taking the time to think deeply about these issues in Teaching 100, I began to wonder whether centering tools like retrieval practice in classrooms (from an early age) and explicitly reinforcing the notion that all meaningful learning is difficult (and therefore must be undertaken iteratively) might boost many people's development of growth mindsets. (Note: it's possible that such a transition in standard educational practice is already somewhat under way, given what my mom said about her experience getting a teaching license several years ago. After all, it's been two decades since I started kindergarten.)
A third aspect of Teaching 100 that I found especially compelling was active learning. The definition of 'active learning' is very broad, but generally refers to educational activities with more student engagement than just listening to an instructor lecture for 60-90 minutes. While I previously had a somewhat lukewarm perception of the push for active learning (especially in higher education), in Teaching 100 we were presented with evidence that in addition to helping a class of students overall, active learning tends to disproportionately benefit students from marginalized racial and socioeconomic backgrounds. The more I reflect on this, the more it makes sense to me that increasing interaction between students and instructors, encouraging and guiding metacognition (self-reflection), and providing some additional course structure can help narrow educational achievement gaps between those who have received more academic / professional mentorship and those who haven't. Deploying active learning strategies to address inequity in the classroom complements my discussion in Reflections on Privilege in Higher Education.
Key Skills Common to Teaching and Consulting
Given my field of study, my thinking about consulting / collaboration is inevitably tied to statistics / data science. However, my sense is that many of these ideas apply much more generally. In fact, reflecting on my experience working to create curriculum for the statistical consulting course, I notice a lot of similarities between best practices taught in Statistical Collaboration and in Teaching 100. I share the most prominent here.
An easy place to start is the importance of cultivating a welcoming environment where people feel respected and are comfortable asking questions and voicing concerns. This is a core tenet both of learning-centered teaching and of building a productive working relationship with a domain expert / statistical consulting client. I won't go into details here, but I believe in the power of both verbal and non-verbal cues!
Two related practices are an immediate focus on identifying each party's goals / motivations and periodically ensuring that the work is in alignment with the goals. As a statistician, this includes inquiring about what the intended product(s) and audience of a domain expert's study will be, and highlighting any mismatches (as they arise) between the question of interest and available data / methods. As a teacher, this includes clearly articulating the course objectives and learning goals (e.g. in a syllabus) and helping students to align their learning experience with their educational / career goals, for instance through intermittent metacognition / reflection assignments. In both settings, writing down a clear plan at the beginning of the partnership, explicitly laying out expectations and timelines for both productivity and communication between parties, is extremely helpful for getting everyone on the same page and making progress towards both parties' goals.
In addition to clarifying overall goals, a communication strategy that resonates a lot with me for both settings is leading with intent. Whether asking a domain expert a question about their data or assigning students a task, giving a brief explanation of why it matters / what you're looking for often yields better outcomes (e.g., information or students' work).
Especially in statistical consulting settings, leading with intent can be used to gracefully assess a domain expert's knowledge of statistics / coding, which then facilitates giving explanations and advice tailored to their level of understanding. In teaching, this is referred to as the zone of proximal development: information and skills that a learner can realistically assimilate into their mental framework in the near term, given what they already know. When starting work in a new area of application, a statistician also must learn on the fly. A technique that we practiced frequently in Statistical Collaboration was to pause after a domain expert's explanation and, instead of immediately moving on to the next thing, summarize back our understanding of what they said. This helps to catch and rectify misconceptions of both the goals and technical details of a project early on, before it becomes burdensome to correct course. Similarly, asking a domain expert or a student to explain out loud (or in writing) what they took away from one of my explanations is a great way to check whether they understood.
In general, a big takeaway for me from both Statistical Collaboration and Teaching 100 is that working with other people is so much more effective and enjoyable when one is familiar with well-tested strategies for doing so. Of course, there is always a chance of unforeseen circumstances that require ad hoc responses. But from my experience and countless stories from others, it seems like a large amount of professional stress is predictable and avoided by consistently implementing existing strategies.
Broad Musings about Modern Education
Continuing on the theme of interpersonal skills, in Teaching 100 our main project was to design a syllabus for a course, and I noticed that many of us in the class chose topics meant to address the "soft side" of our technical disciplines, such as statisticians working with domain experts or doctors working with patients. Traditionally, at least in STEM, there hasn't been much formal training along these lines. Skills in professional communication and/or teaching are often undervalued (e.g. in academic tenure review) and it is mostly expected that people either mimic their advisors / managers or figure it out on their own through trial and error. Especially given the self-selection of many people who have natural affinity for social interaction into other fields, it makes sense that classroom educators might try to fill this gap more systematically.
However, as medicine in particular seems to have known for a long time, professional interpersonal skills are very difficult to learn in only a classroom setting. But outside the classroom, it is harder for educators to actually observe and guide their students' improvement. If the instructor of a statistical consulting course doesn't have the time / compensation to sit through client meetings with every one of their students, then they (and their students) must have a certain amount of confidence in the students' skills to venture forth and report back. This becomes less realistic the more students are in the class and/or the more variation in the students' skill levels. I have yet to arrive at a conclusion about how best to deal with this tension.
Recent events appear to have made interpersonal skills more important than ever. The emergence of tools such as ChatGPT largely level the playing field for many writing-centric tasks, including composition of cover letters. My suspicion is that live interviews and collaboration skills will become ever more important in hiring for professional positions, in addition to promotions.