Why AI Must Show Its Work
The Invisible Learner
During my violin lessons, my teacher often stood behind me. Her hands corrected my posture; her voice guided my bow arm. She couldn’t see my fingers clearly from where she stood, but she could hear the result: the thin sound, the wavering harmonics, the sudden clarity when everything aligned. She taught by listening to what the instrument revealed about my process. The angle of attack, pressure on the string, and shifting weight in my fingers were all hidden, audible only in their consequence.
Now, the education scene has changed. Today, I watch students learn with tools that have no sound at all.
They sit with glowing screens, engaged with systems that promise mastery through algorithms. These polished, intuitive, multimodal AI tools gather resources, generate summaries, create mind maps, and quiz with inhuman patience, solving the peripheral problems of learning: gathering, organizing, and testing. Yet, they create a profound silence where a teacher’s ear once listened. The completed quiz, the generated essay, and the tidy mind map - all are shown, but the process is hidden. We see the note played, but not the finger that found it.
This is the central illusion of our new educational age: that learning can be both accelerated and invisible.
As an IB public school educator with 20 years of experience and a curriculum designer, I’ve witnessed tools come and go. Each promised efficiency, personalization, and engagement. But this generation of AI tools presents something different: not just a new method, but a fundamental shift in the relationship between teacher, student, and knowledge. When a student uses an AI tutor, we see only what they produce with it, not how they arrived there. Did they struggle with a concept for thirty minutes before asking for help? Did they blindly accept a generated summary without questioning its gaps? Did they use the tool to avoid the very cognitive struggle that leads to understanding?
The danger isn’t that these tools are ineffective. In many ways, they’re remarkably effective at certain tasks. The danger is that they make the learning process opaque to the very people tasked with guiding it. Teachers become spectators to outcomes rather than companions in discovery.
When describing meaningful learning, I often think of Tara Westover’s description of education: “a process of self-discovery, of developing a sense of yourself and what you think.” In her memoir, Educated, she describes the transformative power of gaining knowledge: it happens in the wrestling, in the wrong turns, in the moments of confusion that precede clarity. However, when AI smooths those edges automatically, it doesn’t just accelerate learning; it potentially bypasses the formation of the learner altogether. If educators can’t see those moments of struggle, we lose our ability to intervene meaningfully. We can’t celebrate a breakthrough we didn’t witness, or address a misconception that never surfaced.
The critique of “designed for the dump” culture applies here: systems that conceal critical externalities, in this case, the cognitive process. AI learning tools produce clean outputs, hiding the essential, messy work of understanding and risking "cognitive planned obsolescence": learners can perform with AI but lack durable understanding.
So what do we need? We need AI use to be more visible.
We need AI tools that show their work.
We need systems that allow educators to see not just what students produce, but how they interact with knowledge. Imagine an AI dashboard that shows:
-
The questions a student asked the AI before arriving at an answer
-
The points where they paused, rewound, or revisited material
-
The concepts they struggled with versus those they grasped quickly
-
The moments they chose to think independently versus requesting assistance
-
The connections they tried to make between ideas, successful or not
Some will say this sounds like surveillance. But it is not. It restores the teacher’s ability to listen to learning. Just as my violin teacher heard how I played, educators should be able to 'hear' the cognitive process through transparent data. AIED tools need this transparency at their core. AI should augment the teacher-student relationship, not replace it or make it opaque. This means designing systems that:
-
Highlight struggle rather than hide it: Flagging areas of difficulty as opportunities for teaching, not as failures to be smoothed over.
-
Show the path, not just the destination: Making the student’s journey through material visible and meaningful.
-
Empower human judgment: Providing educators with insights they can interpret through their knowledge of pedagogy and their particular students.
-
Resist the illusion of mastery: Being honest about what constitutes real understanding versus pattern recognition.
The best learning has always been a conversation between student and material, between teacher and learner, between old understanding and new insight. AI tools that hide the learner’s side of this conversation ultimately impoverish education. They create what Youtuber Justin Sung calls "passengers in their own learning": students carried along by algorithms rather than developing their own navigational skills.
As a musician, I know that the most beautiful performances come not from perfect technical execution alone, but from interpretation, from feeling, from a personal relationship with the music. Education is no different. The most meaningful learning happens in the personal engagement with ideas, in the development of one’s own voice within a tradition of thought.
AI will not replace teachers. But when AI tools lack transparency, they risk making teachers irrelevant to the most meaningful aspects of learning. We must demand that educational technology prioritize not only efficiency and productivity, but real understanding—and not just outcomes, but the development of each learner.
The future of education depends on tools that show their work and illuminate the process itself. Learning has never been just about the correct answers we find, but about the selves we become in the process.
Isabelle Plante
MA Curriculum Development, Veteran Educator
Co-founder of Sage.Education