On this page

Synthesis: Austin, Gulya and Kassorla argue that generative AI has broken the assumption on which assessment taxonomies were built — that a finished academic product evidences the thinking that produced it. Their conceptual analysis traces a movement across four frameworks: Bloom's hierarchical cognition, Fink's relational significant learning, Kassorla's Inverted Bloom's Taxonomy, which starts from an AI-assisted artifact and works backward, and Austin's own UnBlooms™, a non-hierarchical recursive loop that repositions rather than rejects Bloom. The through-line is that recursion, reflection and metacognitive growth, not the ladder, become the real work of learning when answers are instant and understanding is not. The paper is explicitly conceptual: it compares frameworks and their treatment of process, product, cognition, Learner Agency and metacognition, and reports no empirical data.

Key Findings

  • The pyramid is a misreading, and GenAI exposes why it mattered. Bloom's 1956 taxonomy was hierarchical in intent but never drawn as a pyramid; the pyramid image is a later misapplication. That matters practically, because a ladder implies a learner must master each rung before the next — an assumption generative AI breaks, since students can now produce a polished artifact without the underlying understanding.
  • The product-to-proof link is severed. The authors' central claim is that final artifacts can no longer be read as evidence of underlying cognitive effort. They cite the MAGE framework research showing that systems such as ChatGPT can reliably achieve high grades on standard Critical Thinking tasks including analysis, evaluation and application.
  • Inversion, not abandonment. Inverted Bloom's keeps the six stages of the revised taxonomy but reverses their order — Create, Evaluate, Analyze, Apply, Understand, Remember — so that reliance on the tool is highest at the start and the student's agency rises as the sequence proceeds. It was developed for writing-intensive, artifact-centered assignments.
  • UnBlooms generalizes the move. Austin's framework is problem-centered and recursive rather than linear: it meets learners where they are, whether they begin from an AI output, a piece of misinformation or an original idea, and repositions Bloom rather than requiring its rejection. Inverted Bloom's is presented as a specialized instance of UnBlooms for text-heavy contexts, because laboratory and field-based creation in STEM does not map onto a textual product.
  • The comparison is a trajectory, not a menu. The authors position the four frameworks as a movement from hierarchical cognition, through relational learning and AI-era inversion, to recursive judgment-centered Pedagogies and Teaching Strategies — and argue that professional standards and program structures still built rigidly on Bloom's now work against authentic education.
  • Change has a cost the authors name. They are explicit that faculty cannot redesign courses weekly as the technology shifts, and that designing for recursion and reflection means little without institutional funding and time.

Why the old hierarchy stopped working

The paper situates Bloom's Taxonomy in its original purpose — a shared vocabulary for classifying examination questions, drawn from the taxonomies of science — and in the 2001 Anderson and Krathwohl revision, which converted categories to verbs, swapped Evaluation and Create, and added a knowledge dimension. The problem is not that the taxonomy was wrong but that its use presumes a sequence: comprehension precedes application, and creation crowns the order. Generative AI severs the inferential step from product to proof, so a student can now "create" first and comprehend later, or never. The authors' response is not to discard the framework but to ask what a taxonomy must look like when learners routinely begin from machine-generated output.

Two post-AI frameworks, and how they differ

Inverted Bloom's Taxonomy, developed by Kassorla, is a writing-intensive design: the learner starts with a generated artifact and works backward through evaluation, analysis, application, understanding and recall, with the cognitive load shifting from tool to student at each stage. UnBlooms™, developed by Austin, drops the hierarchy altogether in favor of a recursive problem-centered loop, and is offered as the broader frame of which inversion is one instance. The distinction the authors draw is disciplinary as much as theoretical: in fields where the core work is conceptual modeling, problem solving or laboratory practice, there may be no textual product for a student to "unbloom," and a recursive loop fits where a reversed ladder does not.

What this means for practice

  • Start from the artifact, then work backward. Where a task is writing- or artifact-centered, let students generate first and then evaluate, analyze and revise, so the tool's output becomes the object of critique rather than the answer.
  • Design for recursion, not sequence. Build tasks that cycle between producing, interrogating and adapting, rather than assuming a learner must demonstrate lower-order mastery before attempting higher-order work.
  • Make metacognition the assessed object. If the product can be machine-made, the assessable evidence is the questioning, adaptation and ethical direction the learner applies to it.
  • Treat frameworks as starting points. The authors' advice is to revise rather than preserve cherished taxonomies, and to ask what kind of thinkers a program is building rather than which rung a task occupies.
  • Plan for the workload. Course redesign on this scale needs scheduled time and institutional support; the paper is explicit that expecting weekly redesign from faculty is unrealistic.

Limitations

  • This is a conceptual and theoretical analysis: it compares frameworks and their treatment of process, product, agency and metacognition, and reports no empirical data, no course implementation and no measured outcomes.
  • The frameworks discussed are recent and largely untested; UnBlooms™ is presented as a framework, and the authors note that how students engage in "unblooming" practices directly is left to future work.
  • The comparison centers on higher education and on the authors' own disciplinary contexts, and the claim that Inverted Bloom's fits writing-intensive settings better than laboratory or field-based ones is asserted rather than demonstrated.
  • The paper's policy argument — that professional standards and program structures built on Bloom's now hinder authentic education — is a normative claim, not a finding.

Citation

Austin, T., Gulya, J., & Kassorla, M. (2026). Beyond the Pyramid: Bloom's Taxonomy Meets Generative AI. Postdigital Science and Education.

Connected Concepts

Embed this page

Copy the code below to embed this page in a learning management system or other website. The embedded version shows only the page content — it hides the site header, navigation, and footer.