 Apparently we all need to learn how to prompt now.
How to ask ChatGPT the right question. How much context to provide. How to follow up. How to challenge its response. How to ask it to reconsider. How to get closer to what we're actually looking for. Prompt engineering, they call it. (LOL. That'll be my new title. I'm an engineer. The Director of Prompt Engineering) But the more I think about it, the more I believe it's something we've been doing all along. And when I say "we," I mean I hope "we." Prompting isn't just asking questions, of course. It's also giving instructions, providing context, setting constraints, asking someone to show you something or describe an experience. In other words, shaping what comes back by being thoughtful about what you put in. The efferent and the afferent as you'll see later. But questions — and better questions, hopefully — are a pretty good place to start. Because long before large language models, there was Socrates. And for those of us in clinical practice, before there were LLMs, there was taking a history... An athlete walks into the clinic or treatment area. They have a story. Our job is to try to understand it. And often, the best place to start isn't with a specific question. It's with an open one. "Tell me what's going on." or "How can I help you?" or "Tell me your story." Then listen. Because the moment we start rapid firing or narrowing down our questions, we start narrowing the range of possible answers. Think: Open questions for the history, closed questions for the physical. Not: Does your knee hurt when you run? Yes. Does it hurt going downstairs? Sometimes. Does it hurt when you squat? No. Useful information, maybe. But this doesn't really get us anywhere. At least at the start, I don't think. Compare that with: Tell me about your knee. or, What's that experience like? or, So then what...? We don't necessarily know what we're going to get back. That's the point. It's the athlete's (or patient's) experience, not ours. Key word - experience. In coaching it's no different. A “how did it go?” or “how was that” will get you a “fine.” But a “tell me about that rep” will likely get you a lot more. Sure, sometimes a "good?" is sufficient, but open ended prompts, strategically and intermittently utilized, will likely turn the coach-athlete discourse from 1080p into 4k. Sometimes the best questions aren't even questions. Back to the clinic however, we (hopefully) learned in school that open ended questions create space for the athlete/patient to tell us something we didn't know enough to ask about. Then we can start narrowing. But the later we narrow, the better. Start broad. "Tell me your story." Listen. Then go deep. "Was it more like this (insert characterization) or more like this (insert another characterization) or was it something completely different?" Then come back out again. In-out. In-out. It's a dance. Give and take. Afferent-efferent if you will. It's called an "intake" after all. You, the clinician, takes it in. Let them do the talking. History taking shouldn't be a checklist. I know they teach us that in school (LODRFICARA) but it should be more like an interview. A podcast. And we all know the best podcast hosts are the ones that ask the best questions then let their guests do the talking. They're great at prompting. Based on curiosity. Ask. Listen. Unpack. Ask again. Sometimes go deeper. Sometimes zoom out. Sometimes ask something that, initially, doesn't really have anything to do with the problem sitting in front of you. At least they - the patient/athlete - wouldn't think so. An athlete comes in with knee pain. Ask about the knee and you'll learn about the knee. But ask about what he's been up to in the weeks leading up to the knee "experience" and you may learn that running volume doubled. Perhaps he was trying to catch up after coming back from vacation. Remember "don't bring a manual solution to a load problem?" Or, "first think like a coach than (and then) like a therapist." All of this as they pertain to training injuries. Anyway, ask what else changed around the same time and you might hear about sleep, work, stress, travel, footwear, illness, or a new training program. Truth be told this isn't easy at first, but the more you think this way, the better you'll be able to "prompt." The same athlete, the same knee. Different prompt, different picture. And potentially, a very different understanding of how they got there. Why? Children ask questions. A lot of them. Why? Why? Why? Or The Five Whys. It's not about trying to get to the answer as quickly as possible. It's about trying to understand. Like a kid. Yet many of us seem to be more interested in having answers than asking questions. Ugh. The Five Whys attempts to preserve some of that childlike mentality. Why did this happen? And why did that happen? But why did that happen? And so on. Just keep moving upstream. Not because five is the magical number, but because the first answer is rarely the whole answer. I'd say never. Upstream. Downstream. The key to success in MSK. Think upstream when assessing. When taking a history. Truly figure out the why. Then consider downstream when intervening. Michel Foucault. Randy Travis. Michel Foucault: "People know what they do; frequently they know why they do what they do; but what they don't know is what what they do does." Randy Travis: "I guess it's not what you take when you leave this world behind you / It's what you leave behind you when you go." I'm rambling but again, think upstream to consider the many possible reasons why something is the way it is. Then think downstream to consider the many possible consequences of what we choose to do about it. Curious about causation. Considerate of intervention. Broad. Deep. Broad Again. Deep when necessary. The Neo-Generalist. It's the art of history taking. The art of the prompt. Start broad. Let them talk. Listen. Something catches your attention. Go deep and unpack. Then come back up and look around again. Alice in Wonderland? Sometimes going deeper gives us the answer. Sometimes it gives us a better question. Again, the Neo-Generalist. Broad enough to see the landscape. Deep enough to understand the details. Curious enough to move between the two. But.... What Aren't We Asking? Perhaps this is the more important part. What are we missing? What haven't they told us? What haven't we given them the opportunity to tell us? I'll often "end" my history taking with, "So what do you think you need from me?" I also have a sign up that says, "The greatest gift you can give your child is the opportunity to tell their story." (Note: if you're a parent, let your child do the talking!) Are we asking questions to learn? Or are we asking questions to confirm what we already believe? Admittedly, sometimes we are? Differential diagnoses. Confirm or deny. The art of the exam. But at the end of the day, it's about remaining open-minded. Seeking the truth. Asking the why? Then asking the why again. And again. Which brings us back to AI. An LLM (Claude, ChatGPT, whatever...your choice) is given a prompt. It responds. You give it a narrow prompt, you usually get a narrow response. Pretty much an answer constrained by the boundaries you created. But open the prompt and now you get, or should I say create, possibilities you may not have considered. Then you narrow. And add context. Then go deeper. Sometimes even challenge the response. Ask "what am I missing?" Ask what else might explain it. Ask it to consider another perspective. Look upstream. Consider all the considerations that you may not be considering! (lol). Then think downstream. Ok, so what happens next? And if we do this, then what? Zoom in, zoom out. Dance. Left foot, right foot. Afferent-efferent. Start broad, go deep. Go this way, then go that way. The technology is new. And there will be even newer LLMs that come out. But the thinking isn't. We clinicians and coaches have been prompting for a long time. At least the good ones have been. A clinical history is a prompt. A post-rep debrief is a prompt. A conversation is a prompt. If you're curious, you're a prompter. A Prompt Engineer. The Director of Prompt Engineering. And what I think you'll find is this...Cross-Pollination. Love that word! Cross-Pollination: Making seemingly random connections across disciplines. Because the answer is often found in the question you weren't even thinking about. And because sometimes the thing that changes how we think about rehabilitation comes from places outside of strength and conditioning. From psychology. From engineering. From business. Or somewhere else. So perhaps "prompt engineering" is simply the newest manifestation of something that's been around since Hippocrates. History taking. Podcast interviewing. The first date. Questions. Instructions. Context. Listening. Knowing when to stay broad and when to narrow. The quality of what we get back from an athlete, a patient, a conversation, or an LLM is often determined by the quality of what we put in. It's an art. And maybe it's not about AI after all. Maybe it's better questions. Better prompting. The Art of the Prompt. Stay curious. Comments are closed.
|