Why Doctoral Training Needs Explicit Apprenticeship (Full Transcript)

[00:00:00] Speaker 1: If you clicked on this video, you might think that the PhD system is broken. And as a professor, I'll tell you, yes, it is. But not for the reasons you might expect. You might be surprised why. So in this video, I'm going to look at the core critiques of the PhD system, the assumptions that it's built on, and I'm going to argue why those assumptions no longer hold today, and what that means for you as researchers and the future of the PhD system as we know it. So let's dive in. A lot of the critiques that you see floating around on LinkedIn, Twitter, social media, tend to focus on real problems that are linked to outcomes. There's too many PhDs being trained for too few academic jobs. There's long and variable completion times. And a lot of the skills being cultivated don't translate cleanly outside of academia, and in fact, may be perceived as even being detrimental on the job market. All those are real and serious critiques and problems, but they're downstream of the broken system. They're not really the cause. So I don't want to take away from issues like low pay, inadequate job markets, deep structural inequities as well. They all matter. But they don't get to the heart at what's actually going wrong inside the PhD today. To see that, we need to look at the training model itself. And the core assumption of the traditional PhD is apprenticeship. Think of working away in an ivory tower, and potentially sitting at the feet of a guru who is going to show you what to do. And really, in this model, this idea is that you're learning a craft from a mentor and an expert who is passing the baton onto you. Now, this apprenticeship model that is implicit in a PhD depends on two critical assumptions. One is that supervision is frequent and feedback loops are short. And this is important because this is how you learn to do research. You get shown, not just told, what to do. So frequent supervision means that you get to see how decisions are made. You get to see the standards being demonstrated. You get to see real thinking and logic and judgment applied under uncertainty, not just described in an abstract sense. Short feedback loops mean that errors are caught early. That when there's friction, you learn why something doesn't work. And judgment gets calibrated gradually. It's not like a shift from zero to one, a black to white. And so I often say the way to think about doing a PhD is like getting a driver's license. And in driving terms, the apprenticeship model means you're in the car with the instructor. And the instructor is sitting alongside you. And you might have, when you got your driver's license, you had the instructor had a break next to them. So what that means is they can correct mistakes as they happen. And you learn when to slow down, when to merge, when to abort, when a turn is dangerous, and they can slam on the brake. They also know where you're going to be driving. So it's not as though you step in the car and you're charting the route and map for yourself. Often the driver's instructor is telling you where to turn and telling you where to go at the beginning. That is radically different from what I see happening today. Our training puts you in the car, but often the instructor is not there. And you might be handed a manual, if one at all, and told to study it carefully. And then you're immediately thrust and asked to drive alone on the highway. So here's the problem. These two conditions for frequent supervision and short feedback loops have systematically been undermined over the past few decades for a few reasons. One, supervision time has collapsed. And I can see many students now who come to me saying they get allotted. They are assigned even a limit. This is especially the case in part-time PhDs. One hour of supervision per month. Just one hour. And part of what's going on is student-faculty ratio, especially for PhDs, have gone the wrong direction. They've risen sharply. And so now with this decline in supervision time, it's a bit like checking in with your driver's instructor once every four weeks who's telling you, hey, just go drive for a while and let's see how it goes. But the learner has no idea what counts for doing it well. Often you're doing research during the PhD for the first time. You don't have your feet. You don't know if your idea is good. You don't know if you've gone far enough in your paper. You don't know when it's good enough. The second structural condition that has shifted over time relates to the faculty themselves. And that is apprenticeship is poorly incentivized. So I experienced this myself as faculty. We're rewarded for grants, publications, citations, not for slow, careful apprenticeship or success of our PhD mentees. Unless a student fits into an existing research pipeline, which more commonly happens in the natural sciences where the researchers might be your hands in a lab, unless they fit into any research pipeline, the problem is that the faculty are misaligned structurally with the time and effort invested into the researchers, not necessarily translating into the things that they're getting rewarded for, which is, again, publications, citations, and grants. So that can lead to minimizing time spent, seeing those students as a burden, giving generic feedback, a tendency to rubber stamp progress rather than deeply vetting it to course correct and error correct. And that can lead researchers to go way too far down the chain. And that's not because the supervisors are bad. It's just that the system makes this the most rational equilibrium. The third issue that applies to both is that just the pace of research has accelerated from where it was a few decades ago. Publishing is faster. The volume of the literature is rising exponentially. Expectations are higher. But that doesn't calibrate with beginners who are in a necessary phase of development where they need slower, tighter feedback, not faster systems. Again, this goes back to the highway analogy. They're being asked to get on the highway in a high-speed research environment with less guidance than ever before. So they're going on faster highways, heavier traffic, and we have fewer instructors. So ultimately, what this creates for a lot of research students is a gap in judgment. In the PhD, students are, if at all, taught methods and tools and techniques. But the judgment needed to apply them in the real world under uncertainty with specificities to a research project is just sort of assumed implicitly. Again, this judgment like, is my topic viable? Is it publishable? Is my paper ready to send or premature? When is it good enough? Is this a real contribution? These are some of the highest stake decisions in a PhD. And they used to... That judgment that was passed on was trained through close supervision and apprenticeship. But now, commonly, researchers are just left to figure it out on their own. And that's naturally where they turn to substitutes for this judgment. They start hunting around on YouTube. Could be how you landed here. They use AI to outsource that judgment. Instead of making progress, they start collecting tools, technology that makes them feel confidence, but isn't real progress. Again, none of these are real substitutes for the judgment that gets trained through regular routine feedback. None of these things can replace your driver instructor. So this raises a question. What do we do about this? And when people see the PhD system is broken, some call to even go so far as to abolish it. I don't think that's the right path. But it's clear that it can't work as a casual, part-time, low-contact arrangement and still claim to meaningfully train researchers today. The PhD system simply was built for a world that no longer exists. And so I argue what's needed is a shift from this implicit apprenticeship model to an explicit one. And this would, I think, be best facilitated by two things. One, to make the PhD more outcome-focused. And so this wouldn't just be the time you spend in your PhD, but clear milestones, explicit standards, publishable outputs. And we're seeing this shift happen with more PhD-by-publication-style approaches, where people produce papers rather than a book-length narrative with many chapters. But publishable outputs that demonstrate capability and really are more connected to what you need anyhow for the job market and connect better to the realities and pressures that faculty are facing. I believe this would create greater alignment, greater clarity, and go closer to an explicit apprenticeship model. But that's not the only shift that's needed. There does need to be structural mentorship. It needs to be built in. Because, look, programs that assign one hour per month is not training the judgment that researchers need. And myself, I didn't figure it out on my own. I had multiple mentors who invested in my success, whom I learned from. Many of their insights have formed the basis for what is now our fast-track mentorship system. But you just can't simulate apprenticeship with paperwork, with theory, with books that people read. So if you want people to do serious research, you need to train them how the decisions are actually made. So in driver's terms, we have to show them how to drive in real conditions, in real traffic, on real roads. And that means systems, structures, and feedback. So just to conclude, if you're doing a PhD yourself and feel stuck, ask yourself this. Is this really about ability? Or are you being trained on a model that no longer fits reality today?