An ongoing dialogue on HIV/AIDS, infectious diseases,
August 6th, 2025
Does the Fact That AI Is Brilliant at Writing “Learning Objectives” Prove They’re Not Really Needed?
Paramount Pictures, 1932.
Recently, I was invited to speak at a primary care conference on a terrific topic: “Can’t Miss Diagnoses in ID for the PCP.”
Love it. So many great examples come to mind — endocarditis, Lemierre syndrome, vertebral osteomyelitis, acute HIV. A wonderful opportunity to teach about the “rare but there” diagnoses hidden among everyday outpatient complaints: fatigue, sore throat, back pain, and “mono,” respectively.
I was all in.
You know what I wasn’t all in on? The inevitable mountain of paperwork that followed. Veterans of academic medicine know what’s coming: the dreaded email with five attachments, multiple forms, and of course — the request for Learning Objectives.
First, let me remind everyone of the Universal Rule of Email Avoidance:
The likelihood of an email being answered promptly is inversely proportional to the square of the number of attachments.
Now, back to the Learning Objectives, which will occupy the remainder of this post. They’re short statements (usually three, range 2–5) that describe what the audience is supposed to learn from a lecture. Education experts say they help focus the talk and make evaluation easier. Accrediting bodies like ACCME require them for Continuing Medical Education (CME) credit. (Mea culpa: This is true even for our course, ID in Primary Care, despite this rant to follow. Sorry.) Hence their ubiquity, and the fact that every talk now asks for them.
Some excellent educators defend them as a helpful way to structure a talk. Fair enough; if they work for you, terrific. And there are numerous authorities in both written and video form on how to write effective Learning Objectives, so you can study up to your heart’s content.
(Go ahead, watch one of those videos in its entirety. I dare you. To me, they are mind-numbingly boring — but to each their own, I guess.)
But let’s not pretend Learning Objectives are an essential part of effective teaching. So much of what will be conveyed in a talk can be communicated in the talk’s title, plus (if requested) a 2–3 sentence summary. And that process of using learning objectives to organize a talk? We used to call that “making an outline.” No pre-talk submission required, no adhering to SMART criteria — which stands for “specific, measurable, achievable, relevant, and timebound,” in case you were wondering.
Fortunately, there’s now a solution: Artificial Intelligence — which has gotten so good at writing Learning Objectives that the task is now completely frictionless. I guess this is no surprise, as even in the stone age of AI (like, last year), it was great at streamlining workplace drudgery.
Here’s the prompt:
“Write 3 learning objectives for a talk called ‘Can’t Miss Diagnoses in ID for the Primary Care Provider.’”
Now watch the magic, example #1 from ChatGPT:
At the end of this presentation, learners will be able to:
Identify key clinical features of uncommon but serious infectious diseases that may initially present with nonspecific symptoms in primary care settings.
Apply diagnostic reasoning to distinguish these conditions from more common, benign illnesses through targeted history, physical examination, and appropriate use of laboratory and imaging studies.
Initiate timely management and specialist referral when these high-stakes infections are suspected, to reduce morbidity and improve patient outcomes.
Bravo. These are far better than anything I could have done, mostly because my Learning Objectives are always written with a combination of mild annoyance and the driving motivation not to let “perfect” be the enemy of “just get it done already.”
Didn’t like these ChatGPT products? Go ahead and try Gemini. Or CoPilot.
Voilà! Problem solved. These are so good, it’s downright scary. No thinking required. Copy, paste, submit. Or just send the organizers a link or generate a QR code for handy scanning.
But here’s the problem: if AI can generate them instantly — without context or understanding or board certification in Infectious Diseases — what are Learning Objectives really worth? How many people actually read them? At the end of a talk, no one says, “Wow, those Learning Objectives were beautifully met.”
Instead, we say:
Was that a good talk? Did I learn something useful? Was I engaged? Or was I bored out of my mind, checking my phone?
So I’m sticking with the view that it’s hardly ever about the Learning Objectives. Let’s stop pretending otherwise.
And a plea to the leaders of medical education: could you please — please — lighten the paperwork load for clinical teachers? That would be hugely appreciated.
Let’s close with a scene that proves exasperation with institutional nonsense in education is timeless. These guys had it pegged nearly a century ago.
11 Responses to “Does the Fact That AI Is Brilliant at Writing “Learning Objectives” Prove They’re Not Really Needed?”
Leave a Reply
Paul E. Sax, MD
Contributing Editor
NEJM Journal Watch
Infectious Diseases
Biography | Disclosures | Summaries
Learn more about HIV and ID Observations.
Not all who teach are teachers. In my experience, many doctors who accept an academic position at a medical university wish primarily to pursue their research interests, and teach only because they are required to. This is all too obvious to most medical students. 55 years ago at UNC the dreaded end of year teacher evaluations recognized that reality. All doctors who taught students were required to take a “How to teach” course.
The next year my clinical gynecology professor acknowledge the problem and told my group what he had learned.
“Today we are going to talk about three things, A and B and C. After we are done, you will know how to deal with A and B and C.”
The teaching requirements then as now do not make the person who teaches a good teacher, but make it harder to teach really badly. They are rather like a medical treatment algorithm, outcome oriented. SMART? Uniformity, they say, leads to better results. But it does not necessarily make a better doctor. The doctor who was honored with the teaching award year after year at UNC, Jim Bryant, was a general internist, with no esoteric or even plebeian research interests. He loved to practice medicine, loved his patients, loved his students, and loved to teach. He was a real doctor and oustanding teacher. May we all be.
I know it’s not a big deal, but one thing that really bothers me are attachments that are difficult to fill out electronically, many that require a “signature”. I guess some people have assistants to do this for them, but most of us don’t.
Brilliant blog post, Dr. Sax. And I’d love to hear (or read) your presentation on “Can’t Miss ID Diagnoses for PCPs”.
I love your blog, Dr. Sax, but actually, the ACCME doesn’t have any accreditation criteria that require learning objectives. You can find more information about this here: https://accme.org/news/debunking-cme-planning-myths-tips-for-simplification
Also, the educators (not the learners) are the ones who are expected to evaluate whether the objectives were “beautifully met.” Those outcome data are then supposed (at least in theory) to be used to help plan future CME activities. However, this process doesn’t always happen in the academic CME world, so what you are left with is just the burden of writing the learning objectives and wondering why we actually need them!
Pretty obvious to patients, too, when doctors see them as fodder for their algorithm. Treat the patient in front of you? That is so Jurassic era.
Yes… and… where can I watch that “Can’t Miss Diagnosis” presentation ? !
Has anyone ever shown that demanding Learning Objectives lead to more effective teaching?
AI can also come up with great reasons why “learning objectives at the start of every lecture or rotation aren’t strictly necessary for great teaching”:
• Expert teachers already frame learning in context. The most engaging clinical teaching often unfolds through real cases, questions, and evolving discussion rather than a pre-declared bullet list.
• Adult learners self-direct. Medical trainees come with their own goals and will seek out the details most relevant to their gaps, regardless of what’s written on a slide.
• Medicine is unpredictable. Clinical priorities shift with patient needs, so rigid objectives can constrain opportunistic, high-yield teaching moments.
• Retention is built through experience, not headings. Learners tend to remember the story, the patient, and the reasoning process—not the three “objectives” listed at 8 a.m.
• Objectives can become performative. When they’re a compliance checkbox, they add bureaucracy without improving depth or clarity.
Great teaching in medicine thrives on clarity of purpose, but that clarity can be woven into the teaching itself—without stopping to declare “By the end of this session, you will…”
Ha ha, really enjoyed reading your blog, Dr. Sax, it’s spot on!
Hi! Thanks a lot for the post. It is as interesting as usual. Although I share your concern about utility of “conventional” learning objectives, I do not agree with the main point. If this is so easy for AI, how can this be useful? This completely converts in “no useful” many many things…
cmlnp50@gmail.com
Hello,
In the past few months, I have unable to view the scene at the end of your blog. It always made me smile/laugh. I am not super tech savvy but have checked on my end to see if it might be my problem.
Could you assist me with this as it may be something on your end?
Thank you!