{"id":11593,"date":"2025-08-06T14:59:16","date_gmt":"2025-08-06T18:59:16","guid":{"rendered":"https:\/\/blogs.nejm.org\/hiv-id-observations\/?p=11593"},"modified":"2025-08-06T21:30:14","modified_gmt":"2025-08-07T01:30:14","slug":"does-the-fact-that-ai-is-brilliant-at-writing-learning-objectives-prove-theyre-not-really-needed","status":"publish","type":"post","link":"https:\/\/blogs.nejm.org\/hiv-id-observations\/index.php\/does-the-fact-that-ai-is-brilliant-at-writing-learning-objectives-prove-theyre-not-really-needed\/2025\/08\/06\/","title":{"rendered":"Does the Fact That AI Is Brilliant at Writing \u201cLearning Objectives\u201d Prove They’re Not Really Needed?"},"content":{"rendered":"
\"\"

Paramount Pictures, 1932.<\/p><\/div>\n

Recently, I was invited to speak at a primary care conference on a terrific topic: \u201cCan\u2019t Miss Diagnoses in ID for the PCP.\u201d<\/p>\n

Love it. So many great examples come to mind \u2014 endocarditis, Lemierre syndrome, vertebral osteomyelitis, acute HIV. A wonderful opportunity to teach about the \u201crare but there\u201d diagnoses hidden among everyday outpatient complaints: fatigue, sore throat, back pain, and \u201cmono,\u201d respectively.<\/p>\n

I was all in.<\/p>\n

You know what I wasn\u2019t all in on? The inevitable mountain of paperwork that followed<\/a>. Veterans of academic medicine know what\u2019s coming: the dreaded email with five attachments, multiple forms, and of course — the request for Learning Objectives.<\/em><\/p>\n

First, let me remind everyone of the Universal Rule of Email Avoidance:<\/p>\n

The likelihood of an email being answered promptly is inversely proportional to the square of the number of attachments.<\/em><\/p><\/blockquote>\n

Now, back to the Learning Objectives, which will occupy the remainder of this post. They\u2019re short statements (usually three, range 2\u20135) 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<\/em> them for Continuing Medical Education (CME) credit. (Mea culpa:\u00a0 This is true even for our course, ID in Primary Care<\/a>, despite this rant to follow. Sorry.) Hence their ubiquity, and the fact that every talk now asks for them.<\/p>\n

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<\/a> and video form<\/a> on how to write effective Learning Objectives, so you can study up to your heart’s content.<\/p>\n

(Go ahead, watch one of those videos in its entirety. I dare you. <\/em>To me, they are mind-numbingly boring — but to each their own, I guess.)<\/p>\n

But let\u2019s 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\u20133 sentence summary. And that process of using learning objectives to organize a talk? We used to call that \u201cmaking an outline.”\u00a0 No pre-talk submission required, no adhering to SMART criteria — which stands for “specific, measurable, achievable, relevant, and timebound,” in case you were wondering.<\/p>\n

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.<\/p>\n

Here\u2019s the prompt:<\/p>\n

\u201cWrite 3 learning objectives for a talk called \u2018Can\u2019t Miss Diagnoses in ID for the Primary Care Provider.\u2019\u201d<\/p>\n

Now watch the magic, example #1 from ChatGPT<\/a>:<\/p>\n

At the end of this presentation, learners will be able to:<\/p>\n

    \n
  1. \n

    Identify key clinical features<\/strong> of uncommon but serious infectious diseases that may initially present with nonspecific symptoms in primary care settings.<\/p>\n<\/li>\n

  2. \n

    Apply diagnostic reasoning<\/strong> to distinguish these conditions from more common, benign illnesses through targeted history, physical examination, and appropriate use of laboratory and imaging studies.<\/p>\n<\/li>\n

  3. \n

    Initiate timely management<\/strong> and specialist referral when these high-stakes infections are suspected, to reduce morbidity and improve patient outcomes.<\/p>\n<\/li>\n<\/ol>\n<\/blockquote>\n

    Bravo. These are far<\/em> 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.”<\/p>\n

    Didn’t like these ChatGPT products? Go ahead and try\u00a0Gemini<\/a>. Or\u00a0CoPilot<\/a>.<\/p>\n

    Voil\u00e0! 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.<\/p>\n

    But here\u2019s 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, \u201cWow, those Learning Objectives were beautifully met.\u201d<\/p>\n

    Instead, we say:<\/p>\n

    Was that a good talk? Did I learn something useful? Was I engaged? Or was I bored out of my mind, checking my phone?<\/p><\/blockquote>\n

    So I\u2019m sticking with the view that it’s hardly ever about the Learning Objectives. Let\u2019s stop pretending otherwise.<\/p>\n

    And a plea to the leaders of medical education: could you please — please<\/em> — lighten the paperwork load for clinical teachers? That would be hugely appreciated.<\/p>\n

    Let\u2019s close with a scene that proves exasperation with institutional nonsense in education is timeless. These guys had it pegged nearly a century ago.<\/p>\n