Since the public launch of ChatGPT in November 2022, rapid advances in generative artificial intelligence are reshaping the teaching and learning landscape. Tools that instantly generate text, images, and other products have created an environment where thinking and creativity can be easily outsourced to machines, often leaving educators questioning the authenticity of their students’ work and wondering which cognitive skills are most important to them as learners and performers. It is natural for teachers to be concerned about our future as creative, independent thinkers and problem solvers.
Bloom’s Taxonomy has long been used as a tool that educators can use to identify the level of cognitive demand in their classrooms. Originally developed in 1956 and revised in 2002, this framework provides educators with a common language for curriculum and assessment design. It organizes learning from low-order to high-order thinking skills, starting with basic skills such as memorization and understanding, and progressing to increasingly complex skills such as application, analysis, evaluation, and creation.
But the growing presence of generative AI raises important questions for educators. Does this hierarchy still reflect the mental skills that teachers should cultivate in their students, and perhaps should we abandon Bloom’s framework altogether?
Before the advent of generative AI, creationIntegrating ideas from one’s own knowledge and experience into a final product was considered the pinnacle of cognitive complexity. Now, human authors can create text, images, video, code, or data analysis almost instantly, just by receiving an effective prompt. Creation occurs early in the process rather than as a final step.
In fact, the traditional model of moving from lower-order thinking skills to higher-order thinking skills It doesn’t match the way today’s learners interact with generative AI. Students move flexibly between levels, reflecting on previous content and generating new iterations through additional prompts.
In a generative AI environment, the most difficult cognitive tasks are deciding what to ask, how to frame the question, when to trust or question the output, and how to integrate the AI-generated content into the original work. This type of thinking includes planning (designing clear prompts, setting constraints, and anticipating errors), monitoring (checking output for accuracy, bias, and relevance), and evaluation (criticizing output and modifying prompts). When human-machine collaboration works well, students become active decision-makers in their learning, balancing human reasoning and AI assistance to produce meaningful outcomes.
The traditional model of moving from lower-order thinking skills to higher-order thinking skills does not match the way today’s learners interact with genAI.
This orchestration shares similarities with traditional revision and collaboration, in which learners constantly move fluidly between creating, evaluating, and refining, and reveals that Bloom’s climb through the hierarchy was not a complete picture of how learning actually works. However, AI assistance comes with unique challenges. The speed, scale, and black-box nature of AI-generated content makes metacognitive oversight both more important and more difficult than in human collaboration, as students must manage collaborators who can instantly create sophisticated works without revealing why.
In this new world, skills to remember and understand will be a continuing prerequisite. Learners repeatedly use factual and conceptual knowledge to confirm facts and integrate information throughout the creation and evaluation cycle.
A vertical spiral is better than a pyramid to illustrate relationships between cognitive skills in a generative AI context. This spiral represents a continuous cycle of judgment, modification, and synthesis as learners develop expertise in both content and human-AI collaboration. Learners perform the stages repeatedly, increasing complexity and accuracy with each iteration.
To see how this works in practice, consider the following classroom scenario. A7th Elementary school students researching the subway are required to write an argument for why Harriet Tubman should be featured in a new museum exhibit. He begins by reviewing (remembering/understanding) his notes and primary sources from class about Tubman’s life, the dangers she faced, and the influence she had on others. He writes the first outline and draft. He then creates a prompt and asks his assistant to review his work, citing the assignment’s requirements. “Review this draft argument for why Harriet Tubman deserves to be featured in a museum exhibit about the subway. The assignment requires three reasons supported by historical facts. Does my draft meet these requirements? What historical details could I add?”
The AI generates feedback and suggestions (create), but when students analyze its output (evaluate/analyze), they discover that the AI contains factual errors about the number of slaves Tubman helped free and the reward offered for her capture (recall/understand). He revised the prompt to read, “Help me strengthen my three reasons with specific facts about the number of trips Tubman took, the number of people she freed, and the actual reward amount. If necessary, help me strengthen my use of persuasive voice.”
Students apply their feedback to a revised draft that includes precise details of Tubman’s role in emancipating the slaves. He interweaves his own arguments, AI-based fact corrections and suggested improvements, direct quotes from Tubman that students have independently discovered, and his personal reflections.
As generative AI becomes central to students’ futures, educators must strike a balance between helping students develop strong foundational skills independent of AI, while also preparing students to work effectively with these tools. This requires an intentional teaching strategy that challenges students to build competency without using AI first. Then move on to strategic human-AI collaboration to assess, question, refine, and integrate AI assistance into your own reasoning and unique work.
Although Bloom’s taxonomy still provides educators with a framework for thinking about cognitive demands, this model can better reflect the reality of learning in generative AI environments. The answer is not to abandon Bloom completely, but to rethink it to emphasize an iterative learning cycle of judgment, critique, and synthesis.
Teachers who accept this reality must design tasks that make thinking visible by requiring students to evaluate output, identify errors and biases, refine prompts, and integrate AI assistance with their own reasoning. When we equip our students with both traditional competencies and AI literacy, we can prepare them not as passive consumers of technology, but as skilled directors of this human-machine collaboration. This is exactly what they want for their future.
