The Abstracting-Contextualizing Model of Mentorship

If you've ever been told to find a theory for your project and walked away more confused than when you started, this post is for you. MedEdMentor started about six years ago when Greg and I struggled to explain the processes essential to health professions education (HPE) scholarship. We knew theories and frameworks were important, but we didn’t know how someone should most efficiently go about finding and applying them to their work. We sought to tackle this from two angles. The first was to provide resources to make this task easier and more straightforward. And the second was to learn what experienced mentors do to help novice HPE scholars find and use theories.

As you likely guessed, our approach to providing resources to novice scholars is what you have come to know as MedEdMentor. We started by creating an interactive database of as many theories as we could find that were relevant to HPE. That narrow focus has since expanded dramatically, as our understanding of the needs in HPE scholarship has grown, leading to what you all find today.

Now, we haven’t spoken much about our work trying to understand how experienced mentors support novice scholars to find and use theory…until now. Published ahead of print last month in Academic Medicine is the culmination of this work: The Abstracting-Contextualizing Model (ACM) of Mentoring for Theory Use in Medical Education.

Over the course of six years, 20 interviews, and countless hours reading and re-reading (and re-reading) those transcripts, we learned innumerable nuggets of wisdom, all coalescing around the ACM. I would love it if you read the whole paper (email me if you are having trouble accessing it), because I would love to hear what you all think of it. However, as a teaser, the ACM in its simplest form is described below.

Problems

Problems are the specific and tangible problems in our everyday educational lives that we aim to solve, like a new workshop on responsible use of ChatGPT in clinical settings that the internal medicine (IM) clerkship director asked you to build to deliver to medical students that are preparing to enter their IM rotations.

Phenomena

Phenomena are the generalized versions of problems. Something that anyone in HPE could recognize and imagine in their own context. In this case, the phenomenon might be: principles of responsible use of commercially available LLMs by learners in clinical settings. ChatGPT, medical students, internal medicine, and clerkships are all gone. Now, what you are focused on is specific enough to be meaningful, but general enough that you can learn from, and share what you learn with, colleagues in other fields. And not just across medical disciplines, like pediatrics, surgery, dermatology, etc. I’m talking about nursing, pharmacy, dentistry, and veterinary medicine, etc.

Theories

Before we talk about theories and frameworks, I need to state, I don’t differentiate between these terms and will use “theories” as a catchall phrase going forward. Theories are tools and approaches that support improved understanding of phenomena. Take our phenomenon from the previous paragraph. We can see what's already in the literature on LLM use by learners in clinical settings. But what are we fundamentally interested in? Maybe it's how learners adopt new technologies, in which case, there are theories from outside HPE that apply. Maybe it's the ethics of presenting work that isn't fully your own. Say a medical student uses a hospital-approved, HIPAA-compliant LLM to help turn a patient note they wrote into a coherent oral presentation. Do they need to disclose that before they present? Different ethical frameworks give us different answers. The possibilities are endless!

The Abstracting-Contextualizing Model

The actual Abstracting-Contextualizing Model is a system of techniques and approaches that experienced mentors use to help their mentees to navigate between these three key components. They first identify where their mentees are starting their scholarly journey, whether they come with a concrete problem, interest in understanding a general phenomenon, or a theory they think would be useful to HPE. They then facilitate abstracting from a specific problem to its analogous general phenomenon, as illustrated in the example outlined above. Next, they serve as “theory brokers,” proposing various tools and lenses that their mentees could use to better understand their phenomenon, and helping them to explore each one as they look for a match that feels right to them.

A simplified schematic of The ACM. The full model can be found in the manuscript.

The flip case is just as important. Imagine a scholar trained in cognitive psychology with deep expertise in cognitive load theory. They want to apply it to HPE. Their mentor doesn't push them away from theory; they help them find the HPE phenomenon their expertise helps to illuminate. That might well be the same one we've been circling: LLM use by learners in clinical settings. Cognitive load theory opens up a productive set of questions here. Does the LLM lessen learners' cognitive load enough to make learning easier? Does it lessen it so much that they don't really learn? Or does it actually increase the load and become a hindrance?

These are just some of the insights you'll find in the full paper, and they're also what shape how MedEdMentor works. The ACM serves as the skeleton upon which MedEdMentor is being built to serve you the way an experienced HPE mentor would. The HPE paper database we've built serves as the interstitium, filling in the intercostal spaces. Hahaha, sorry, I'm taking the anatomy analogy too far.

When you use MedEdMentor, you're not using a generic commercial LLM. You're using a tool built on the wisdom of experienced mentors and backed by empirical evidence.

This was the hardest single project I have ever worked on, and I learned more from doing it than from anything else I've done academically. If you read the paper, I'd genuinely like to know what you think, especially where you'd push back. Please, email me your thoughts at geoffstetson@mededmentor.org. That's how the next version of this work, and future iterations of MedEdMentor, will get better.