Human Judgment in an AI World

It does not matter what conference I attend or what group I am speaking with, everyone is talking about artificial intelligence (AI).

We have accepted that AI is not coming. It is here.

The questions now tend to come quickly:

• How can we use AI in the classroom?
• How can AI help teachers?
• How can students use AI responsibly?
• How can schools prepare for what comes next?

Those are understandable questions, but I am beginning to think we are spending too much time asking how.

“How might we use this?” is a brainstorming question.

It invites possibilities, encourages experimentation, and helps people imagine new applications.

However, when a tool can generate a plausible response to almost any prompt within seconds, possibility is no longer the hardest part.

The harder question is this:

How do we develop the judgment to decide which plausible responses should be accepted, revised, challenged, or rejected entirely?

That is not merely a technology question, it is an education question.

Plausible Is Not the Same as True

One of the most useful things to understand about generative AI is that it is very good at producing responses that sound reasonable.

The language may be polished, structure may be convincing, explanation may even appear complete. It may also produce a lesson plan that looks thoughtful or a realistic image. Any recommendations it produces may sound authoritative.

None of this means the response is accurate, appropriate, useful, ethical, or even connected to reality.

It is only plausible.

This distinction matters enormously.

A plausible response can resemble knowledge without actually containing much knowledge, and certainly no wisdom. 

A plausible response can give the appearance of expertise without demonstrating understanding.

A plausible response can produce a polished artifact that collapses the moment someone asks a deeper question.

Sometimes what looks like a finished building is really just a convincing facade. Scratch the surface, and there may be nothing holding it up.

This is where human judgment becomes essential.

AI May Get You Most of the Way There

People often say that AI can get you 80 percent of the way to a finished product.

That may be true in some situations. AI is certainly helpful in generating a first draft, organizing ideas, offering examples or suggesting structure. 

Generative AI can help someone move more efficiently from a blank page to something that can be evaluated.

However, that remaining 20 percent is not necessarily a minor finishing step.

Sometimes that 20 percent is the part that determines whether the work is actually good.

It is where someone asks:

• Is this accurate?
• Does this make sense?
• Does it fit the purpose?
• What is missing?
• What assumptions are embedded here?
• Whose experience has been ignored?
• Would this work in the real world?
• What would happen if we acted on this?

Those questions require judgment from people capable of thinking critically. 

Without judgement, we are not using AI to produce meaningful work.
We are simply producing more content, more text, more images, more lesson plans, more recommendations.

All of these are more artifacts that look complete but may not actually move the real work to be accomplished forward in any meaningful way.

Efficiency is only valuable when it actually helps us accomplish something truly worthwhile.

Producing the wrong thing faster is not progress.

The Most Important Skill May Be Rejection

Much of the conversation about AI has focused on what people can create with it. This makes sense because the generative capabilities are impressive. 

However, I think one of the most important skills in an AI-rich world may be the ability to evaluate and reject what has been generated:

• No, that conclusion is not supported.
• No, that explanation is too simplistic.
• No, that recommendation ignores important context.
• No, this sounds good, but it is not actually useful or realistic.
• No, this is not the right tool for this particular problem because people are not widgets.

That kind of discernment requires more than just technical skill. Rather, it requires the knowledge of content and awareness of context that are typically earned through experience.

Underpinning all of this is the requirement of ethical reasoning.

“Ethical reasoning is the ability to identify, assess, and develop ethical arguments from a variety of ethical positions.”

Wesleyan University

Critical Thinking Has Always Been the Goal

In some ways, this is not a new educational challenge.

What is new are the tools, the speed and the scale. The underlying need to develop critical thinking has always been there.

“critical thinking…(a) mode of cognition using deliberative reasoning and impartial scrutiny of information to arrive at a possible solution to a problem.”

Brittanica

K–12 education has never been only about helping students acquire information.

Reading matters, but students must also learn to interpret what they read.

Writing matters, but students must also learn to decide what is worth saying and whether their reasoning is sound.

Mathematics matters, but students must also learn when and how to apply quantitative reasoning.

Knowledge matters, but education is not complete when students can merely repeat information.

The deeper goal is to help young people develop the capacity to think.

• To weigh evidence.
• To recognize weak reasoning.
• To ask better questions.
• To distinguish confidence from competence.
• To consider consequences.
• To revise their thinking when new information becomes available.
• To make decisions in situations where the answer is not immediately obvious.

These questions together form the basis of the act of critical thinking, which is how one exercises human judgment. In a world filled with tools capable of generating plausible responses, that capacity becomes more important, not less.

Students Learn What Adults Normalize

This is where the conversation becomes more uncomfortable.

Adults in educational systems often talk about how students should use AI.

We discuss student rules, student integrity, disclosure, and dependence. While these conversations matter, we cannot ignore that students are also watching how adults use these tools.

Students see whether:

• teachers accept generated materials without examining them closely.
• leaders circulate summaries they have not verified.
• polished language is treated as evidence of thoughtful work.
• speed is valued more than accuracy.
• remain responsible for the things they produce, regardless of which tool helped produce them.

What adults model eventually becomes what students experience and expect as normal.

• If adults use AI as a shortcut around thinking, students will learn that the purpose of the tool is to avoid thinking.
• If adults accept plausible responses without scrutiny, students will learn that plausibility is enough.
• If adults treat generated output as finished work, students will learn that completion matters more than understanding.

This responsibility places a greater burden on adults, not a smaller one. The question is not only whether students are using AI responsibly, we have to ask if the adults around them are demonstrating what responsible use actually looks like.

Before we tell students what is responsible use of AI, the adults teaching and leading them have to adhere to those same expectations.

We Cannot Teach Discernment We Do Not Practice

Educational systems cannot ask students to evaluate AI-generated content critically while adults are using the same tools uncritically.

We cannot tell students to verify sources if we do not verify them ourselves.

We cannot insist that students explain their reasoning while accepting generated recommendations that no one in the room can explain or defend.

We cannot teach students that tools should serve a clear purpose while adopting every new platform because it promises efficiency, personalization, innovation, or transformation.

Students learn discernment partly through explicit instruction, but also by watching adults decide when a tool is useful, revise generated work, question confident-sounding answers, acknowledge uncertainty, and remain accountable for the final decision. 

That modeling is part of the curriculum whether we intend it to be or not.

Human Judgment Needs Good Information

Judgment does not happen in a vacuum.

People cannot make strong decisions without access to relevant, accurate, and appropriately interpreted information.

This is one reason the issue matters so much to me as a measurement professional.

Assessment, at its best, provides information that supports better judgment. It does not replace judgment. 

• A test score cannot decide what a student needs next.
• A dashboard cannot determine whether an initiative is working.
• A performance level cannot explain the full experience of a learner.
• Data can inform.
• Evidence can clarify.
• Measurement can reveal patterns.

People still have to decide what the information means, what it does not mean, and what action is justified.

AI belongs in that same category. It can support judgment, increase efficiency, surface possibilities, and assist people to begin or further a body of work. What it cannot do is relieve us of the responsibility to think.

The more powerful the tool becomes, the more disciplined the human judgment surrounding it must become as well.

The Right Question Is Not Whether AI Is Good

Conversations about technology often become unnecessarily binary.

• Is AI good or bad?
• Should schools embrace it or resist it?
• Will it improve learning or destroy it?
• Will it help teachers or eliminate them?

Those questions may attract attention, but they are not especially useful.

Tools are rarely good or bad. Every technological advancement is more or less appropriate for particular purposes.

A tool may be excellent for generating examples and terrible for making a consequential decision. It may be useful for organizing ideas and unreliable for verifying facts. It may help a teacher draft a family communication and be inappropriate for evaluating a student’s needs.

The better questions are:

• What problem are we trying to solve?
• What does this tool actually do well?
• What risks does it introduce?
• What knowledge must the user already possess to evaluate the result?
• What human judgment must remain involved?
• What would responsible use look like in this situation?

Those questions are less exciting than asking what AI might eventually do, but they are also far more important.

Education Must Develop the Capacity to Judge

Access to public education represents an extraordinary commitment.

As a society, we have decided that children should have access to years of learning during the period in which they are becoming adults. That commitment is not only about workforce preparation, or even about teaching reading, writing, and arithmetic.

It is about preparing people to participate in a society that depends on people being able to think, reason, evaluate information, and make decisions. An informed citizenry requires more than access to information, it requires the ability to judge information. This has always been true.

AI simply makes the consequences of failing to develop that capacity much more visible.

We are entering a world in which answers are easier to generate than judgement, and while plausibility is abundant, wisdom is not. 

No matter how quickly content can be created, understanding still takes time.

K–12 education has a critical role to play here by intentionally helping students develop the capacity to examine, challenge, apply, revise, and sometimes reject what a powerful tool places in front of them.

Human Judgment Still Lives in the Real World

AI operates within its systems, its training, its prompts, and its limitations.

People live in the real world.

We live with the consequences of our decisions, the complexity of human relationships, and the contradictions that arise every day. We recognize when context changes the meaning of an answer. Most importantly, we carry ethical responsibility for the decisions we make.

This is why human judgment is indispensable.

Some will respond by avoiding AI altogether. Others will embrace it uncritically. I believe a better path is to develop people who can use these tools without surrendering their responsibility to think.

In a world filled with AI, I believe the most important question is not whether we can use it, but whether we have developed the discernment to know what to do with it.

If that is the world our students are inheriting, then developing human judgment may be one of the most important responsibilities education has ever had.