AI Should Support Teacher Judgment, Not Replace It

AI, Teachers, and the Human Work of Schools — Part 2

In the first post of this series, I wrote about the new expectation creep surrounding artificial intelligence in public schools — the way AI is being introduced with urgency, promise, and pressure, while many teachers are still being left to learn it on their own time.

That problem matters because time is not a small thing in teaching. Time is where professional learning happens. Time is where ethical questions are worked through.  Time is where teachers begin to understand not only how a tool works, but whether it belongs in the life of a classroom.

This second post turns to an even deeper question.

Who gets to decide?

As AI becomes more visible in lesson planning, grading, feedback, tutoring, assessment, intervention, and curriculum design, schools will need to be very clear about the role teachers are meant to play. Are teachers the professional decision-makers who use AI thoughtfully as one possible support? Or are they gradually being moved into the role of implementers, expected to manage tools and systems designed somewhere else?

That distinction matters.

Because AI may be able to generate a lesson plan. It may be able to draft a rubric. It may be able to suggest comments on a piece of writing. It may be able to summarize data, organize resources, and produce materials quickly.

But it does not know the child.

It does not know the classroom.

It does not know the quiet look on a student’s face when the assignment is too hard but the student is too embarrassed to ask for help. It does not know the history behind a parent email, the emotional texture of a class period, the student who is capable of more than the data suggests, or the child who needs encouragement more than correction.

Teaching has never been only the delivery of content.

It is the practice of judgment in the presence of human beings.

The Risk of De-professionalization

Every major school initiative carries a hidden question: Does this strengthen the profession, or does it weaken it?

AI is no different.

Used well, AI could support teachers by reducing clerical burden, helping generate first drafts, organizing information, or giving teachers more time for the relational work that only humans can do. That possibility should not be dismissed. Teachers are overloaded, and tools that reduce unnecessary work deserve thoughtful attention.

But there is another possibility.

AI could also be used to standardize more of teachers’ work, automate decisions that require professional discernment, or quietly shift authority away from educators and toward platforms, vendors, algorithms, and policy systems.

That is the danger of de-professionalization.

De-professionalization does not usually happen all at once. It happens gradually. A tool is first introduced as a support. Then it becomes an expectation. Then it becomes embedded in a workflow. Then teachers are asked why they are not using it. Then the tool begins shaping what counts as efficient, effective, aligned, or acceptable.

Over time, the teacher’s judgment can become secondary to the system’s output.

That should concern us.

Not because teachers are opposed to technology, but because teaching requires forms of knowledge that cannot be reduced to a prompt, a dashboard, or an automatically generated recommendation.

Teachers interpret context. They adjust in the moment. They know when the lesson plan needs to be abandoned. They know when a student’s silence is confusion, resistance, fatigue, fear, or grief. They know when an answer is technically wrong but developmentally promising. They know when feedback should push and when it should protect.

Those decisions are not inefficiencies.

They are the work.

AI Can Draft, But Teachers Must Decide

One of the most important distinctions in AI policy should be the difference between assistance and authority.

AI can assist.

Teachers must decide.

– AI can draft a lesson objective, but teachers must decide whether it fits the students in front of them.

– AI can generate sample feedback, but teachers must decide whether that feedback is accurate, humane, developmentally appropriate, and likely to help the student grow.

– AI can summarize patterns in student work, but teachers must decide what those patterns mean and what instruction should come next.

– AI can suggest a reading passage, but teachers must decide whether the passage is culturally responsive, accessible, rigorous, accurate, and aligned with the deeper purpose of the lesson.

– AI can help with administrative tasks, but it should not become the hidden author of professional judgment.

This is where policy needs to be precise. It is not enough for districts to say teachers may use AI responsibly. Policies need to define where AI can support professional work and where human review is required. They need to make clear that AI-generated content is not automatically instructional quality. They need to protect teachers from being pressured to accept system-generated recommendations that conflict with their professional knowledge of students.

The U.S. Department of Education has emphasized the importance of keeping humans in the loop when AI is used in teaching and learning. That is the right starting point. But in schools, “human in the loop” cannot mean a teacher merely clicking approve on something produced elsewhere. It has to mean real authority, real discretion, and real professional responsibility.

The teacher should not be the loop.

The teacher should remain the professional center.

The Human Knowledge of Teaching

One of the reasons AI conversations often feel incomplete is that they underestimate the kind of knowledge teaching requires.

Teaching is not simply knowing content. It is not simply knowing methods. It is not simply managing a classroom. It is the integration of subject matter, pedagogy, child development, culture, timing, relationship, motivation, emotion, and judgment.

A teacher does not just ask, “What should I teach?”

A teacher asks, often in the same moment:

• What do these students already understand?
• Who is ready to be pushed?
• Who needs more safety before they can take a risk?
• What misconception is underneath that answer?
• What is the mood of the room today?
• What does this student’s behavior mean in context?
• What will help them believe they can try again?

That kind of knowledge is not generic.

It is embodied, relational, and situated. It grows through years of practice, reflection, failure, revision, and care. It is part craft, part science, part moral attention.

This is why AI should be treated as an aid to professional judgment, not a substitute for it.

A teacher might use AI to generate several versions of a writing prompt. But the teacher knows which version will invite deeper thinking from this particular class. A teacher might use AI to create a draft parent communication. But the teacher knows the family, the history, the tone, and the relationship. A teacher might use AI to suggest scaffolds for a struggling student. But the teacher knows whether the student needs a scaffold, a challenge, a conversation, or simply a little more time.

The tool may provide options.

The teacher provides wisdom.

Voluntary Use and Professional Autonomy

One policy question districts should take seriously is whether AI use should be voluntary, required, or limited to specific approved purposes.

There may be contexts where certain AI-related practices become part of district systems, especially around administrative efficiency or approved instructional supports. But when AI tools affect lesson design, feedback, grading, student interaction, or classroom routines, teacher autonomy must be protected.

Not every teacher will need the same tool. Not every subject will benefit in the same way. Not every grade level raises the same questions. Not every student population faces the same risks. And not every AI use that looks efficient from a distance will be instructionally wise up close, with real people – students.

A humane AI policy should avoid turning experimentation into compliance.

Teachers should not feel pressured to use AI simply because it is new, promoted, purchased, or embedded in a district platform. Nor should they be penalized for exercising professional caution. In the early stages of AI integration, thoughtful restraint may be as responsible as thoughtful use, perhaps more so.

Voluntary-use clauses can help protect this balance. They can clarify that teachers may use approved tools when those tools support instructional goals, but they are not required to replace their professional methods with AI-generated materials or processes. Such language respects both innovation and professional judgment.

It also prevents AI from becoming another performative expectation: one more thing teachers feel they must display in order to appear current.

The goal should not be AI use for its own sake.

The goal should be better teaching, deeper learning, and healthier conditions for the people doing the work.

Teacher-Led AI Advisory Committees

If districts are serious about responsible AI integration, teachers should not merely be trained after decisions are made. They should help make the decisions.

Every district considering AI tools should create a teacher-led AI advisory committee or working group. This group should include educators from different grade levels, subject areas, school contexts, and student support roles. It should also include specialists in special education, multilingual learner support, counseling, library/media services, and technology integration.

The purpose of such a committee would not be symbolic input. It would be substantive guidance.

Teachers should help evaluate which tools are worth piloting, which tools create more work than they save, which uses raise privacy or equity concerns, and which professional learning needs to accompany implementation. They should help define appropriate use by age level and subject area. They should help identify where AI might reduce clerical burdens and where it might interfere with student thinking, teacher-student relationships, or authentic assessment.

This kind of committee also creates a healthier feedback loop.

Instead of AI policy being written at the top and interpreted differently in each building, teacher-led advisory structures allow districts to learn from the people closest to the work. That improves trust. It also improves policy.

Teachers know where the friction is.

Teachers know where the risk is.

Teachers know where the promise might actually be.

If we want AI to serve schools well, we should begin by listening to the people who understand schools from the inside.

AI and the Temptation of Efficiency

One reason AI is attractive to school systems is that it promises efficiency.

That is understandable. Public schools are under enormous pressure. Administrators are asked to do more with less. Teachers are overloaded. Students need more support. Budgets are often tight. Families want responsiveness. Policymakers want results.

In that environment, any tool that promises speed will get attention.

But efficiency is not the highest value in education.

Some parts of teaching should not be rushed. Some conversations need time. Some feedback needs care. Some mistakes need to be understood before they are corrected. Some students need relationship before they can receive instruction. Some learning cannot be optimized without being thinned out.

The danger is that AI may tempt systems to treat slowness as failure.

But much of the human work of schools is slow.

A teacher waiting long enough for a student to find their words is slow. A teacher reading between the lines of a rough draft is slow. A teacher noticing that a student’s performance has changed over the last two weeks is slow. A teacher adapting a lesson because the room is not ready for what was planned is slow.

That slowness is not waste.

It is attention.

A humane AI policy must be careful not to replace attention with output. 

If AI helps reduce paperwork so teachers have more time for attention, then it may serve the human work of schools. But if AI becomes a way to demand more output from teachers, more data from students, more standardized feedback, more monitored performance, and more compliance disguised as innovation, then it has not strengthened education.

It has simply accelerated the wrong things.

Equity and the Human Role

The question of teacher judgment is also an equity question.

AI systems can reflect bias. They can produce inaccurate information. They can misunderstand dialect, culture, disability, multilingual development, or context. They can generate materials that appear polished but are not pedagogically sound. They can recommend interventions without understanding the full life of a student.

Students who are already vulnerable are often most affected when systems substitute procedural efficiency for human discernment.

This is why teacher judgment is not a luxury. It is a safeguard.

A teacher who knows a student can question the data. A teacher who understands cultural context can revise the material. A teacher who sees the whole child can resist a recommendation that reduces that child to a category. A teacher who has built trust can notice what an algorithm cannot.

If AI is used in under-resourced schools primarily as a shortcut — a way to compensate for staffing shortages, reduce human interaction, or provide cheaper forms of support — then it risks widening the very inequities public education is supposed to address.

The students with the greatest needs should not receive the least human attention.

AI may help teachers serve students more effectively, but only if it is used to strengthen the teacher’s capacity, not replace the teacher’s presence.

What Policy Should Say Clearly

A strong AI policy should not leave teacher authority implied. It should name it directly.

– It should say that teachers remain the final professional decision-makers in matters of instruction, feedback, grading, classroom use, and student support. 

– It should require human review of AI-generated instructional materials before they are used with students. 

– It should prohibit AI from making final decisions about grades, placement, discipline, services, or student evaluation. 

– It should clarify that AI tools may support teacher work, but they do not replace teacher responsibility or professional judgment.

– It should also protect teachers from unreasonable expectations.

If a district encourages AI use, it must provide training, time, approved tools, privacy guidance, and ongoing support. If a district pilots an AI product, teachers should have a meaningful voice in evaluating its impact. If a tool increases workload, reduces autonomy, or produces weak instructional materials, teachers should be able to say so without being labeled resistant.

Good policy should create a shared understanding:

AI may assist the teacher.

AI may not become the teacher.

AI may support judgment.

AI may not replace judgment.

That boundary matters.

A Better Vision

The best future for AI in schools is not a future where teachers are replaced by technology, nor is it a future where teachers reject every new tool out of fear.

The better future is one where teachers are trusted enough to use tools wisely.

In that future, AI handles some of the repetitive tasks that drain energy from the school day. It helps draft, organize, translate, brainstorm, and adapt. It gives teachers starting points, not final answers. It reduces some of the clerical load so teachers have more time for students.

But the teacher remains the one who decides what matters.

• The teacher listens.
• The teacher notices.
• The teacher interprets.
• The teacher adjusts.
• The teacher protects the dignity of the learner.
• The teacher holds the human story behind the data.

That is not a sentimental view of teaching.

It is a professional one.

Public schools do not need policies that make teachers feel like supervisors of machines. They need policies that honor teachers as skilled professionals whose judgment is essential to any ethical use of technology.

AI may be part of the future of education.

But the future of education still depends on human beings who know how to teach, how to care, how to discern, and how to remain present with students in all their complexity.

AI should support that work.

It should never replace it.

Resources