What if Bloom’s Taxonomy is no longer enough for the AI era?
Posted by: Strayo De Agarwal
I recently came across Tina Austin’s thought-provoking article, The Evolution of Bloom’s Taxonomy, and the accompanying UnBlooms™ Framework. It sparked a question I believe medical education needs to grapple with more deliberately.
For decades, medical educators have relied on Bloom’s Taxonomy to structure learning objectives and educational design:
Remember. Understand. Apply. Analyse. Evaluate. Create.
It’s elegant. Familiar. Foundational.
But generative AI has changed the equation.
Today, AI can assist at almost every level of the pyramid. It can retrieve information faster than we can, summarise evidence, generate differential diagnoses, draft management approaches, analyse data, and increasingly critique and refine its own responses.
So when a learner demonstrates competency, a difficult question emerges:
Did the learner do the thinking, or did the AI?
This is what struck me about the UnBlooms™ Framework: the idea that learning may be better represented not as a ladder, but as a loop:
Question → Generate → Critique → Refine
Non-hierarchical. Recursive. Context-dependent.
You can enter the cycle at any point. What matters is not where you start, but whether you are actually doing the cognitive work.
And in many ways, this mirrors clinical reasoning itself. Clinicians rarely move through a neat linear hierarchy. We generate hypotheses, challenge assumptions, revisit evidence, refine decisions and adapt to new information, often cycling through these steps repeatedly.
Now consider that through the lens of IME/CME.
We often design learning as a clean progression: recall the pharmacokinetics, apply the knowledge to a case, analyse the trial data, evaluate competing regimens, develop a treatment plan.
But an HCP today can generate a differential diagnosis with a prompt, summarise a landmark trial in seconds, or produce a polished case-based answer without necessarily wrestling with the uncertainty and ambiguity that build clinical judgement.
The output may look like mastery. The learning underneath it may not have happened.
And this raises a bigger question:
What exactly are we assessing when an HCP produces the “right” answer?
For years, many educational assessments have focused on what learners can produce at a particular moment. But in an AI-mediated learning environment, a polished output is becoming a much weaker signal of whether meaningful, durable learning has actually occurred.
For IME/CME providers, I see this as both a challenge and an opportunity.
The emphasis shifts from what was produced to how the learner arrived at it.
Did they question the recommendation? Did they recognise an assumption? Did they identify what was missing? Did they challenge the evidence?
Perhaps the future isn’t about simply adding an AI component to an existing educational format.
Perhaps it is about designing AI-aware learning experiences in which the learner’s interaction with AI becomes part of the educational process, and potentially part of what we measure.
Perhaps this is where medical education needs to evolve: from measuring whether an HCP can produce an answer to understanding whether they can exercise sound clinical judgement in an environment where answers are increasingly easy to produce.
I’m curious to hear from fellow educators, faculty, clinicians and learning scientists:
As AI becomes embedded across the learning journey, should IME/CME focus less on measuring answers and more on measuring judgement, discernment and reasoning? How are you thinking about designing learning and assessment in an AI-mediated environment?
________________________________________________________________________
Credit to Tina Austin and the creators of the UnBlooms™ Framework for provoking this conversation.