[AI × Learning] From Mental Autopilot to Active Thinking: Why You Must Learn to Ask Before Using AI

Introduction

In an age overflowing with information, a question aimed at a real goal is evidence that you have not given up thinking.

Does This Sound Like You?

Hong-An opened ChatGPT and, after thinking for a long time, typed an opening request: “Here is my original ad. Please rewrite it in a brand voice that feels gentle but firm.”

Hong-An expected a rhythm and direction he had never considered. Instead, ChatGPT’s response felt like the original copy had simply been rearranged.

“Was my description of the tone too vague? Or did I ask the wrong question?” Hong-An wondered.

Now consider Hsiao-Yin. She wanted to use Copilot for her final book-reflection assignment, so she typed: “Please write a 500-word reflection on the book Your Kindness Needs an Edge.”

Copilot promptly produced 500 words. Hsiao-Yin had finished the assignment quickly, but she also felt that something of her own participation was missing. She tried again: “Can you make it sound more like me?”

Copilot produced another long response. Something had changed, yet it still felt wrong. “Did I ask badly? Am I actually thinking?” she wondered—then decided to submit it anyway.

The scene shifts again, this time to Pei-Cheng, a new employee who had never used AI. An hour earlier, her manager had asked her to handle a supplementary amusement-tax filing and told her to ask AI if she did not know how.

After staring at the chat box for a long time, Pei-Cheng typed: “How do I file amusement tax?”

Gemini listed documents to prepare and channels for filing. Yet the answer did not solve her problem: she did not need to make an ordinary filing; she needed to make a supplementary one. She was still completely lost.

Does any of this sound like the way you use AI?

What Can Go Wrong When AI Is Used Poorly

A recent MIT Media Lab study of large language models (LLMs)_[1] divided participants into three groups: an LLM group that wrote with ChatGPT, a search-engine group that could use tools such as Google, and a brain-only group that used no tools. The study involved four writing sessions with crossover conditions and examined brain activity, natural-language patterns, and interview responses.

In the LLM group, as many as 83% of participants could not accurately quote from the essays they had just completed. This result concerns verbatim quotation performance in the study—not general memory ability. The researchers also observed the weakest brain connectivity, the lowest sense of ownership over the resulting text, and greater similarity in vocabulary and themes across participants in the LLM group._[1]

In this specific essay-writing task, the groups using external tools showed less cognitive engagement, with the LLM group displaying the weakest brain connectivity. The finding is a useful warning: repeatedly handing comprehension, judgment, and memory work to a tool may save effort in the moment while reducing active participation. However, this small preprint study cannot establish that every form of AI use causes long-term cognitive harm._[1]

The authors use the term “cognitive debt” for a possibility worth watching: when people repeatedly complete tasks at a lower cognitive cost, the effort saved in the short term may be followed by weaker memory, understanding, and autonomous engagement. This remains an interpretation and warning arising from one study design; larger samples and different tasks are still needed._[1]

In a Nature report on the same study, Nicola Jones noted that participants who first wrote with an LLM and then switched to writing without tools still showed lower brain engagement in the fourth session._[2] The report also cautioned against overinterpreting a small experiment with a limited sample. It does not support the claim that reliance on AI causes an irreversible decline in brain activity.

In a separate editorial, Danish psychiatrist Søren Dinesen Østergaard reviewed emerging cases involving generative-AI chatbots and delusions._[3] He warned that fluent mimicry and interactions that readily affirm a user could pose risks to people already vulnerable to psychotic symptoms. At present, however, the evidence consists mainly of cases and risk hypotheses. It does not establish that long-term use of generative AI generally causes or intensifies delusions; systematic research is still needed.

Taken together, these warning signs suggest that if all we ever do is wait for AI to answer without thinking, we may someday worry that AI really will replace us—and start saying “thank you” after every prompt in hopes that it will spare us later.

“If AI is that frightening, should we simply stop using it?” You may be wondering the same thing.

Use Questions to Reopen the Possibility of Thinking

But not every user works this way.

Some people open an AI window with a clear sense of direction. They explain what they need: “I am organizing an issue. Here is what I have so far; please help me…” They can quickly produce text that fits their own style and purpose.

They ask precise follow-up questions, restate the problem to clarify uncertainty, and continue until the task is complete. Then they turn what they learned into teaching materials, reflective writing, or explanations for other people.

They are not using AI merely to obtain an answer. They collaborate with it to complete a task. AI is a partner, but they retain control of the pace and direction.

With AI, they accomplish things that once seemed impossible, and the collaboration can even accelerate their own growth.

This is the mental toolkit used by an active-thinking questioner.

Which Kind of Questioner Are You?

Before continuing with the rest of the series, this article offers a set of questions to help you examine your usual habits and state of mind when asking questions or using AI.

This questionnaire is not a test of whether you are good at asking questions. It helps you notice your current level of voice participation. Each option represents a different conversational rhythm and motivation. Wherever you begin, the rest of the series will help you find a questioning style and pace that work for you.

As you read those questions, you may notice that they are not measuring how often you use AI, how professionally you use it, or whether you know how to operate it.

They are closer to questions for self-examination. When you ask something, are you an observer, a participant, or the person guiding the exchange? What kind of conversational space is your voice trying to create? Are you being carried forward, feeling your way toward a direction, or designing the entire interaction?

The same person may fall into different profiles in different situations. Social and work contexts, for example, may produce completely different results.

Four Questioner Profiles

Theoretical Foundations of the Framework

In designing this questioner framework, we drew on two psychological theories: self-determination theory (SDT)_[4] and expectancy-value theory (EVT)._[5]

These theories approach motivation from the perspectives of basic psychological needs and task engagement. Together, they help us think about the depth of a user’s linguistic participation and their conversational agency when interacting with AI.

Their intersection also offers a more precise way to describe different patterns of voice participation and to design later modules for questioning and learning with AI.

Self-Determination Theory (SDT)

Self-determination theory is a framework for human motivation and personality development. It distinguishes intrinsic motivation from different forms of extrinsic motivation.

Intrinsic motivation means engaging in an activity for its inherent enjoyment, challenge, or satisfaction, without needing an external incentive. Extrinsic motivation means acting to reach an external outcome, such as a reward, recognition, or responsibility.

SDT treats intrinsic motivation as a high-quality form of motivation. When someone begins with little motivation or mainly extrinsic motivation, support for three basic psychological needs can help motivation become more self-determined: autonomy, or experiencing one’s actions as chosen rather than forced; competence, or feeling capable of completing a task and meeting a challenge; and relatedness, or feeling connected, understood, and supported by others.

Ryan and Deci further describe a continuum from less self-determined to more self-determined motivation:

• Amotivation: a complete lack of motivation or intention to act.
• External regulation: acting for rewards, to avoid punishment, or to meet another person’s demand.
• Introjected regulation: acting to avoid guilt, gain self-esteem, or preserve one’s self-image.
• Identified regulation: recognizing the value of a behavior and willingly carrying it out.
• Integrated regulation: bringing a behavior into alignment with one’s values until it becomes part of the self.
• Intrinsic motivation: engaging spontaneously for the enjoyment, challenge, or satisfaction inherent in the activity.

In other words, SDT does not deny the value of extrinsic motivation. It emphasizes how thoroughly that motivation has been internalized and how autonomous it has become. When autonomy, competence, and relatedness are supported, extrinsic motivation can gradually develop into higher-quality, self-determined behavior.

From this perspective, the quality of motivation matters more than its sheer intensity. Durable engagement and psychological satisfaction are more likely when behavior is self-determined. SDT therefore does more than classify motivation: it helps us see that the quality of linguistic participation depends on whether a person’s psychological needs are supported.

Expectancy-Value Theory (EVT)

Expectancy-value approaches trace back to John W. Atkinson’s work in the 1950s and 1960s. In explaining the connection between motivation and behavior, the theory proposes that whether a person undertakes a task depends on their expectancy of success—“Can I do this?”—and the value they assign to it—“Is this worth doing?”

Eccles and colleagues later distinguished several components of subjective task value:

• Attainment value: how important the task is to one’s identity or sense of self.
• Intrinsic value: the enjoyment or interest involved in doing the task.
• Utility value: how useful the task is for future goals.
• Cost: the effort, anxiety, opportunity cost, or other burdens involved.

The article also distinguishes efficacy expectancy—whether I believe I can perform the required behavior—from outcome expectancy—whether completing it will produce the result I want. These are established English terms, although this particular distinction is more closely associated with social-cognitive theory than with the core Eccles expectancy-value model.

A simplified expectancy-value account also suggests that if either expectancy or value approaches zero, overall motivation will tend to approach zero as well.

EVT reminds us that motivation is not a single act of will. It emerges where “I can do this” meets “This is worth doing.”

Emerging Voice User: Participation Has Not Yet Begun, and Motivation Is Still Taking Shape

While talking with AI, you may have said something like this—

“I was simply told to use AI, but it is so complicated that I do not even know where to begin.”

“I do not really understand the differences in tone. As long as it produces something I can submit, that is enough.”

“I pasted someone else’s prompt, but the result does not seem right…”

There is nothing wrong with having these thoughts. They are an early awareness of your own voice in an AI conversation. You are not alone; you are beginning the journey toward conversational agency.

Early on, some users feel that they are only carrying out a task and that AI is not something they can truly control. It can feel like a runaway horse. They may even worry that the AI will think they are foolish.

In motivational terms, this can resemble amotivation, or a situation in which interaction with AI is driven mainly by external pressure—rewards, punishment, or someone else’s demand.

This does not mean the person lacks ability. It may instead show that the three psychological needs relevant to participation—autonomy, competence, and relatedness—are not yet sufficiently supported. SDT suggests that as these needs are met, motivation can be activated and extrinsic motivation can become more self-determined.

EVT likewise reminds us that motivation forms where “I believe I can do this” intersects with “I believe this matters.”

If you are an Emerging Voice User, this series will help you clarify your purpose before asking, learn from opening examples, and find a workable conversational rhythm. The aim is to move from observing the exchange to participating in it—and to build the safety and direction needed to say, “I can start asking.”

Exploring Voice User: An Intention Has Appeared, but the Conversational Rhythm Is Still Developing

Or perhaps you have thought something like this—

“AI can solve this problem, but I am not sure how to ask well.”

“AI answered my question, but something still feels missing, and I do not know how to adjust it.”

“How can I get AI’s response to sound more like the way I speak?”

If these thoughts sound familiar, you are not alone. You have already taken the courageous first step toward guiding your conversations with AI.

Some users have begun to recognize the importance of voice participation. They know they should take part in and guide the exchange, yet AI’s responses can still feel like an unruly rudder. The words may answer the question, but the interaction feels one-sided—or as though both sides are talking past each other.

In a simplified EVT framing, this resembles high value paired with low expectancy: the person knows the task is worth doing but is unsure they can do it. They may also worry that a long back-and-forth with AI will consume time without producing what they need.

From an SDT perspective, these users may already experience some autonomy and some relatedness through communities that treat AI as important. Their motivation to engage with AI and shape the rhythm of the exchange has begun.

If you are an Exploring Voice User, this series will show you how to establish a questioning rhythm, check response quality, and use follow-up questions efficiently. You can move from “I want to ask” to “I know how to ask” and gradually develop your own way of working with AI.

Designing Voice User: Conversational Agency Is Established, and Strategy Building Has Begun

If you regularly use AI to solve problems, you may sometimes wonder—

“I know exactly what I need to solve, but which question should I ask AI first?”

“Have I asked enough? Should I keep going? How can I extend the question?”

“Can AI really help with this problem? How can I verify that?”

Once you become comfortable with AI, it can feel like a small partner you open every workday. It handles tedious tasks, discusses copy with you, and may even prompt you to rethink your workflow.

These users already exercise conversational agency and are beginning to design voice and strategy. They do not merely participate in language; they shape it—adjusting tone, reframing questions, and challenging responses. In the article’s applied framework, this resembles integrated regulation in SDT and a high-expectancy, high-value state in EVT.

For a Designing Voice User, the series offers ways to plan the rhythm of a conversation, evaluate its quality, and improve it through follow-up questions. The goal is to explore rhythm, style, and logic so that AI becomes a better working partner.

Creative Voice User: Style and Mirroring Rhythm Are Stable, Opening the Way to Creation and Challenge

“AI is my co-creator!”

“AI helps me grow and create possibilities I could not reach on my own.”

If you are a Creative Voice User, AI is no longer merely a tool. It is a collaborator with whom you can discuss many kinds of work.

These users have entered a creative stage of language use. Their conversational agency is stable, and their style and mirroring rhythm are highly autonomous. Within the article’s SDT application, their motivation resembles intrinsic motivation, and linguistic participation has become part of self-realization.

Within the article’s EVT application, expectancy and value are strongly integrated. Language is no longer merely a tool; it becomes a creative medium and a means of accelerated learning and growth.

For these users, the series will share situations in which I use AI, the contexts and methods behind my questions, and problems and playful projects I have explored with it. The project is still continuing.

It bears repeating: this series is not meant to rank anyone as better or worse. It is meant to help every user find a workable rhythm and move from observing language to designing it.

Whether you have not begun, feel stuck in uncertainty, or want to go further, there is support for your voice. AI can move from being a rigid tool to becoming a genuinely friendly collaborator.

Move from Using AI to Training Your Thinking

“Knowledge should not simply be poured into us; it should be built through questions, exploration, and construction.” This captures a central idea of inquiry-based learning (IBL),_[6][7] a learner-centered approach to teaching and learning.

In this era, AI is an available partner that never needs to sleep. We can combine an inquiry-based cycle—questioning, investigating, explaining, communicating, and reflecting—with deliberate questioning and conversational agency to learn more effectively.

Instead of merely seeking an answer, we can confirm, deepen, and extend ideas through an exchange with AI, turning the interaction into a process of knowledge construction:

• Questioning: begin with a first question that helps clarify your own thinking.
• Investigating and explaining: compare your understanding with the AI’s response. Ask whether its claims rest on reliable sources and reasoning, and check for logical errors.
• Communicating: after checking your understanding, ask for detail, extend the original question, and continue the exchange until the need or goal is met.
• Reflecting: use AI to examine blind spots in your learning process, identify what you may have missed, and consider better ways to ask and interact.

Combining AI with inquiry-based learning gave rise to this series. This article opens “AI × Learning,” which will address the following themes:

• How can you organize your thinking before asking?
• How can you help AI understand who you are, your voice, and your goal?
• Find a conversational rhythm and questioning style that fit you.
• Use follow-up questions to improve AI’s responses.
• Learn when to conclude the conversation and start acting.
• Attend to the psychological states that surface while you ask.
• Translate these modules into learning tools for situations without AI.
• Practice project: rescuing a hard drive.

The series is not merely about procedures or prompts—although there will be plenty of prompts. It is about rebuilding the habit of thinking through questions and shaping a personal method for learning with AI.

Within this system, you will encounter the following templates and modules:

• Motivation Recognition Module: understand why you are asking.
• Goal Anchor Generator: design a goal anchor suited to your conversational motive.
• Opening Prompt Generator: complete a questionnaire and receive a detailed opening message that can be given directly to AI.
• Conversation Anchoring Method: use a sequence of mechanisms and interaction choices so that AI no longer feels like a runaway horse.
• Quality-Check Workflow: learn to tell whether an AI that seems to hallucinate is actually making things up.
• Semantic Follow-Up Workflow: use designed follow-up techniques to request clarification whenever something remains uncertain.
• Stopping-Point Checklist: recognize when to stop asking and start doing.
• AI-Assisted Learning Workflow: use AI to support accelerated learning and growth.

Someday, when you open an AI tool, you will no longer hesitate over how to begin or worry about choosing Copilot, ChatGPT, or Gemini.

Your first sentence will carry a motive. You will know what you want to explore or co-construct, how your voice should express who you are, and how to adjust the collaboration.

You will no longer simply accept the response. You will be able to say, “I took part in this” and “This reflects how I think.”

You will have rhythm, follow-up questions, a clear ending, and language that cares for your emotional state. You will no longer feel defeated simply because AI did not produce the answer you imagined.

Conversation will stop being only a way to complete tasks and become a daily training ground for your own voice.

This is what completing the system can help you become:

A person with linguistic agency—someone who can use voice to shape thought and use AI to learn and grow. This series is designed to help you reach that point.

Conclusion

You begin to do more than enter text. Each question becomes a way to participate in thinking. However good an AI answer may be, it matters less than learning how to ask.

How do you want to begin today’s conversation? Try one small sentence—just enough to bring yourself into the thinking.

A good question does not prove that you are clever. It proves that you are still thinking. But before you ask—do you really know why you are asking?

In the next chapter, we will help you identify the motivation and purpose behind your questions.

References

[1] Kosmyna, N., Hauptmann, E., Yuan, Y. T., Situ, J., Liao, X.-H., Beresnitzky, A. V., Braunstein, I., & Maes, P. (2025). Your brain on ChatGPT: Accumulation of cognitive debt when using an AI assistant for essay writing task. arXiv. https://arxiv.org/abs/2506.08872

[2] Jones, N. (2025, June 25). Does using ChatGPT change your brain activity? Study sparks debate. Nature. https://doi.org/10.1038/d41586-025-02005-y

[3] Østergaard, S. D. (2025). Generative artificial intelligence chatbots and delusions: From guesswork to emerging cases. Acta Psychiatrica Scandinavica, 152(4), 257–259. https://doi.org/10.1111/acps.70022

[4] Ryan, R. M., & Deci, E. L. (2000). Self-determination theory and the facilitation of intrinsic motivation, social development, and well-being. American Psychologist, 55(1), 68–78.https://doi.org/10.1037/0003-066X.55.1.68

[5] Wigfield, A., Tonks, S., & Klauda, S. L. (2009). Expectancy-value theory. In K. R. Wentzel & A. Wigfield (Eds.), Handbook of motivation at school (pp. 55–75). Routledge. https://doi.org/10.4324/9780203879498

[6] Spronken-Smith, R. (2005). Experiencing the process of knowledge creation: The nature and use of inquiry-based learning in higher education. University of Otago. Retrieved from Ako Aotearoa summary report

[7] Pedaste, M., et al. (2015). Phases of inquiry-based learning: Definitions and the inquiry cycle. Educational Research Review, 14, 47–61. https://doi.org/10.1016/j.edurev.2015.02.003