Prompting Is a Form of Thinking

In the previous series, I focused on the forms of human thinking that become increasingly important in an AI-rich world: judgment, sensemaking, thinking under uncertainty, curiosity, intellectual humility, ethical reasoning, transfer, and metacognition.

The final post focused on metacognition: the ability to examine our own understanding, recognize gaps in our knowledge, evaluate the strategies we are using, and adjust our approach when necessary.

That leads to a practical question. If these are the human capacities, we want students to develop, how can students work with AI in ways that strengthen those capacities rather than weaken them?

That is the focus of this new series: learning alongside AI.

I want to begin with prompting. Prompting is often presented as a technical skill. Students are given formulas, templates, and lists of words that supposedly produce better results. They may be taught to assign AI a role, provide context, specify a format, identify an audience, or include examples.

Those strategies can be useful. A clearer prompt will often produce a more relevant response. But I think prompting is more than a technique for getting better output. Prompting is a form of thinking.

Before students can clearly tell AI what they need, they have to make sense of the task themselves. They have to decide what they are trying to accomplish, identify the information that matters, recognize the constraints, and determine what kind of support would be helpful.

In other words, a prompt can reveal how a student understands the problem.

Consider the difference between these two prompts:

“Write my report about cloud computing.”

And:

“I am comparing public, private, and hybrid cloud models for a mid-sized healthcare organization. Help me identify the criteria I should use to evaluate the three options. Do not recommend one yet. Ask me questions about the organization’s needs before helping me build the comparison.”

The second prompt will probably produce a better response. But the important difference is not simply its length or level of detail.

The student has begun thinking about the problem.

They have identified the context. They recognize that the options should be evaluated according to criteria. They understand that the organization’s needs matter. They have also decided that they are not ready for a final recommendation.

That prompt reflects a different relationship with AI. The first prompt hands over the task. The second uses AI to support a process the student is still directing.

The distinction matters because when a student writes a vague prompt, the problem may not simply be a lack of prompting skill. The student may not yet understand the task well enough to ask for meaningful help.

If a student says, “Tell me about networking,” the request is broad because their understanding may still be broad. They may not know which part of networking is confusing, what prior knowledge they have, or what kind of explanation would help.

That is not necessarily a failure. It may be the beginning of learning. The prompt gives us information about where the student is. A more developed prompt might say:

“I understand that switches connect devices within a local network, but I am confused about how a router’s role is different. Explain the difference using a small-business example, and then ask me two questions to see whether I understand it.”

Now the student has identified what they already know, what remains unclear, and how they want to test their understanding.

The value of the prompt is not only that AI has received better instructions. The student has had to examine their own understanding.

·      What do I already know?

·      What am I uncertain about?

·      What kind of explanation might help?

·      How will I know whether I understand the answer?

Those are metacognitive questions. Prompting can make them visible.

We often teach prompting as if students simply need to add more information. But a long prompt is not automatically a thoughtful prompt.

Students can copy elaborate prompting formulas without understanding the problem any more deeply. They can assign AI a role, specify a tone, request a table, and describe the desired length while still outsourcing the central thinking.

A polished prompt can still say, in effect, “Do this for me.”

The more important question is not whether the prompt contains all the recommended components. It is whether the student has made meaningful decisions about the task.

·      What is the real problem?

·      What information is relevant?

·      What assumptions am I making?

·      What constraints need to be considered?

·      What do I want AI to help me do?

·      What thinking should remain mine?

These questions move prompting beyond a formula. They turn it into problem framing.

That matters because AI will usually respond to the task it is given, even if the task has been defined poorly. It may generate a confident answer to the wrong question. It may accept assumptions that should have been challenged. It may move toward a solution before the student understands the situation.

A well-written response cannot rescue a poorly framed problem. That is why one of the most valuable things students can do before prompting AI is pause.

They can ask themselves what they are actually trying to understand or accomplish. They can identify what they know and what they still need. They can decide whether they need an explanation, an example, a question, a comparison, feedback, or an alternative perspective.

That pause may be more educationally important than the prompt itself.

The role students assign to AI also matters. The same tool can participate in learning in very different ways depending on what a student asks it to do.

A student might ask: “Give me the answer to this troubleshooting scenario.”

Or the student might ask: “Help me work through this troubleshooting scenario. Ask me what I would check first. After I respond, challenge my reasoning and give me one additional piece of information at a time. Do not provide the solution unless I ask for it.”

In the first case, AI removes the problem. In the second, AI helps the student remain inside the problem.

A student writing a paper might ask AI to generate a thesis. Or the student might describe two possible arguments and ask AI to identify the assumptions and weaknesses in each one.

A student studying for an exam might ask AI to summarize an entire chapter. Or the student might explain the chapter in their own words and ask AI to identify what appears to be missing or confused.

A student completing a technical assignment might ask AI to recommend a solution. Or the student might provide a proposed solution and ask AI to generate questions a skeptical supervisor would ask.

None of these approaches guarantees learning. AI may still provide inaccurate, incomplete, or unhelpful responses. Students still need to evaluate what it produces.

But the prompts establish different relationships with the tool. One asks AI to perform the work. The other asks AI to participate in the student’s thinking.

That participation should also be iterative. Prompting is sometimes taught as if the goal is to create one perfect instruction that produces the ideal response. But that is not usually how meaningful thinking works.

We ask a question. We receive information. We notice something we had not considered. We revise our understanding. Then we ask a better question.

Prompting can follow the same process.

A student may begin with a broad question because they are unfamiliar with the subject. The response may introduce language, concepts, or distinctions that help them see the topic more clearly. That should lead to a more focused question.

They might ask for clarification. They might challenge part of the response. They might add a constraint. They might request an example that does not fit the original explanation. They might ask AI to compare two interpretations. They might recognize that the first question was not the question they actually needed to ask. That revision is not evidence that the first prompt failed. It can be evidence that learning occurred.

The student’s questions changed because their understanding changed.

This is one reason I am cautious about presenting prompt engineering as a fixed set of rules. Students do need strategies for communicating clearly with AI, but they also need to understand that prompting is a process shaped by evaluation and revision.

The goal is not to produce the perfect prompt on the first attempt. The goal is to ask increasingly better questions. If prompting is part of the thinking process, then we may also want students to reflect on it rather than allowing it to remain an invisible step.

That does not mean instructors need to collect every conversation a student has with AI. Nor does it mean students should be required to write lengthy explanations after every use.

Small additions can make the thinking more visible. Students might submit the first prompt they used and one revised prompt, then briefly explain what changed. They might identify what information they added after seeing the initial response. They might explain why they asked AI for questions rather than answers. They might describe one assumption in their original prompt that they later reconsidered.

We might ask them:

·      What were you trying to accomplish with this prompt?

·      What did you already understand before asking?

·      What kind of help did you want from AI?

·      How did the response change your next question?

·      What part of the task did you intentionally keep for yourself?

These questions shift attention away from whether a student used the “correct” prompting formula. They focus attention on the decisions behind the prompt.

That is especially important because effective AI use will not look identical in every discipline or situation. A useful prompt for brainstorming will look different from a prompt used for tutoring. A prompt asking for feedback will look different from one designed to generate practice scenarios. A low-risk creative task will not require the same context or caution as a consequential professional decision.

Students do not simply need a universal prompt template. They need to understand what they are asking AI to do and why.

There is nothing wrong with teaching students how to communicate clearly with AI. Clear instructions, appropriate context, and meaningful constraints can improve the quality of a response.

But if prompting instruction focuses only on obtaining better output, we risk missing the educational opportunity.

The deeper value of prompting is that it can require students to define problems, identify gaps, establish criteria, articulate constraints, and decide what kind of support they need.

Those are human thinking skills. They matter even when AI is not involved.

A student who can clearly frame a problem is better prepared to research it, discuss it, solve it, and explain it to someone else. A student who can identify what they do not understand is better prepared to seek useful help. A student who can distinguish between needing an answer and needing a question is becoming a more intentional learner.

That is where prompting connects to the larger purpose of this series.

Learning alongside AI is not simply about becoming more efficient at using a tool. It is about remaining an active participant in the thinking process.

A prompt should not merely tell AI what to produce. It should reflect a decision about how AI will participate in the work.

·      What do I need?

·      What do I already know?

·      What am I trying to learn?

·      What role should AI play?

·      What thinking still needs to remain mine?

When students can answer those questions, prompting becomes more than a technical skill. It becomes a form of metacognition. It becomes a way of framing problems. It becomes a way of making thinking visible. And perhaps most importantly, it becomes a way for students to direct AI without allowing AI to quietly direct them.

But even when students thoughtfully decide how AI should participate, another question remains. What happens when they hand part of the thinking over to the tool? That is where the next post will turn: cognitive offloading and the decisions students need to make about what they should and should not ask AI to do for them.

Continuing the Conversation

Series 1: AI Is Exposing Existing Problems ✓ Completed
Series 2: What We Do About It ✓ Completed
Series 3: Cultivating Human Thinking ✓ Completed
Series 4: Learning Alongside AI
Current Post (1 of 8): Prompting Is a Form of Thinking
Next Up: Cognitive Offloading: What Should We Hand Over to AI?