AI Beyond Prompts: Why I Taught People to Question AI, Not Just Use It

By Ealeonore Mafor

Artificial intelligence is changing how people learn, work, conduct research, make decisions, and solve problems.

Tasks that previously demanded considerable time and mental effort can now be completed with the assistance of AI in a matter of seconds.

But as artificial intelligence becomes more capable and accessible, an important question deserves greater attention.

Are people becoming better thinkers with AI, or are they gradually becoming more dependent on it to do their thinking?

This question became one of the driving forces behind my AI Beyond Prompts Masterclass.

I believe artificial intelligence presents remarkable opportunities for learning, innovation, and productivity. But I am equally concerned about what happens when people begin accepting AI-generated answers without questioning, evaluating, or understanding them.

My objective was not simply to teach people how to use ChatGPT or write better prompts.

I wanted to teach them something more fundamental: how to think with artificial intelligence rather than surrender their thinking to it.

The Problem I Wanted to Address

The growing adoption of artificial intelligence has introduced an interesting contradiction.

Technology is making information easier to access, yet accessing information is not necessarily the same as understanding it.

An individual can ask an AI chatbot to solve a problem, explain a concept, write an assignment, or develop a business strategy without necessarily understanding the reasoning behind the response.

For some users, the process ends once the AI generates an answer.

There is no attempt to question the information, investigate its assumptions, explore alternatives, or determine whether the conclusion is accurate.

Over time, this pattern raises concerns about intellectual dependence.

If people consistently allow AI systems to perform tasks that ordinarily require reasoning, analysis, evaluation, and problem-solving, what happens to their willingness or ability to practise those skills independently?

This is particularly important for students, professionals, entrepreneurs, researchers, and young people developing foundational thinking skills.

The issue is not that using AI is inherently harmful.

The concern is uncritical dependence on AI, especially when it replaces rather than supports cognitive effort.

What Research Says About AI and Critical Thinking

My concern is not based solely on personal observation.

Research has begun examining how increasing reliance on generative AI influences human thinking and cognitive engagement.

A 2025 study by researchers from Microsoft Research and Carnegie Mellon University, titled The Impact of Generative AI on Critical Thinking, surveyed 319 knowledge workers and examined 936 examples of AI-assisted work.

The researchers found that greater confidence in generative AI was associated with less self-reported critical-thinking effort.

Interestingly, greater confidence in one's own abilities was associated with more critical-thinking engagement.

The study also observed that AI can shift people's cognitive effort away from gathering information and performing tasks toward verifying outputs and overseeing the work.

Similarly, a 2024 systematic review published in Smart Learning Environments examined research on students' overreliance on AI dialogue systems. It identified concerns involving critical thinking, analytical reasoning, and decision-making.

Another 2025 research project, Your Brain on ChatGPT, examined how different approaches to essay writing were associated with brain activity and task engagement.

The small experimental study reported lower measured neural connectivity during essay writing among participants using a large language model compared with participants writing without tools.

These findings deserve attention.

However, they should also be interpreted carefully.

The existing evidence does not establish that artificial intelligence is making humanity universally less intelligent or that using AI inevitably causes permanent cognitive decline.

What it does suggest is that the way people engage with AI matters.

When AI replaces opportunities to reason, question, remember, and solve problems independently, there may be cognitive costs.

When used deliberately, however, AI can also support learning, exploration, and more sophisticated problem-solving.

That distinction informed my approach to teaching.

Why I Chose to Teach AI Instead of Discouraging Its Use

The world is already experiencing widespread AI adoption.

Artificial intelligence is becoming integrated into search engines, productivity software, educational platforms, communication tools, and professional workflows.

For many people, avoiding AI completely is neither realistic nor particularly beneficial.

But adopting it without understanding its limitations is not a responsible alternative.

I saw a need for a different approach.

Instead of telling people not to use artificial intelligence, I wanted to teach them how to use it without becoming intellectually dependent on it.

Instead of encouraging people to memorise hundreds of prompts, I wanted them to understand how AI systems behave, why they sometimes produce incorrect information, and how to evaluate their responses.

Instead of teaching AI merely as a tool that receives instructions and completes tasks, I wanted to introduce it as something that could support deeper thinking.

A student should be able to use AI to understand a difficult subject without allowing it to replace the learning process.

An entrepreneur should be able to evaluate an idea with AI without accepting every recommendation.

A professional should be able to work more efficiently without abandoning responsibility for the quality of their work.

And an engineer should be able to use AI during development while still understanding, testing, and defending the technical decisions made.

These were the kinds of capabilities I wanted to encourage through AI Beyond Prompts.

If artificial intelligence is going to become part of how people work and learn, then teaching people how to think critically while using it is increasingly important.

Demystifying Artificial Intelligence

Over the past two years, I have taught more than 1,500 people about artificial intelligence.

Through my experience in AI education, I have observed how easily conversations about the technology become dominated by tools, prompts, and impressive demonstrations.

But AI literacy goes considerably deeper.

Artificial intelligence can appear intimidating to people who do not have technical backgrounds.

Some assume that understanding AI requires advanced programming skills.

Others believe AI systems are so intelligent that their answers should rarely be questioned.

Neither assumption provides a useful foundation.

During the AI Beyond Prompts Masterclass, I wanted to make AI understandable without exaggerating its capabilities.

My approach was to connect technical concepts with everyday situations.

I explored what AI systems can do, where their limitations exist, how they generate responses, and why their outputs require evaluation.

More importantly, I wanted participants to understand that they did not have to become software engineers before developing practical AI literacy.

Artificial intelligence is relevant to far more than the technology industry.

Its applications and consequences affect education, entrepreneurship, communication, research, and everyday decision-making.

Understanding it should therefore be accessible to a much wider audience.

Understanding AI's Sycophantic Nature

One of the most important topics I explained during the masterclass was AI sycophancy.

Sycophancy, in the context of conversational AI, refers to the tendency of a system to agree with or affirm a user's beliefs, opinions, or assumptions, even when a more critical or corrective response would be appropriate.

Imagine presenting a business idea to an AI chatbot.

You explain that the idea is brilliant, that no competitor has developed anything similar, and that you are convinced it will succeed.

An excessively agreeable AI assistant may respond by praising the idea, highlighting its potential, and encouraging you to proceed.

But what if your assumptions are wrong?

What if similar solutions already exist?

What if customers are unwilling to pay?

What if the technology is expensive to develop or the business model is unsustainable?

An AI system that simply agrees with you may reinforce overconfidence rather than improve your judgment.

This is why I taught participants to recognise the difference between agreement and useful analysis.

An AI response should not be considered valuable simply because it tells you what you want to hear.

Sometimes the most useful response is the one that challenges your assumptions.

Turning AI Into a Thinking Partner Rather Than a Command-Taking Assistant

Another major focus of the masterclass was changing how participants approached AI conversations.

Many people use artificial intelligence as a command-and-obey tool.

They enter instructions, receive responses, and move on to the next task.

That approach can be perfectly appropriate for routine activities.

But when the objective involves learning, strategy, reasoning, or important decisions, simply requesting answers may not produce the greatest value.

I encouraged participants to approach AI differently.

Instead of using it exclusively to generate content or confirm opinions, I taught them to use it as a critical thinking partner.

A thinking partner should help you examine ideas, identify weaknesses, explore different perspectives, and challenge unsupported assumptions.

For example, instead of asking:

"Tell me why my business idea is excellent."

I encouraged a more critical approach:

"Evaluate my business idea objectively. Identify my assumptions, challenge weak reasoning, examine potential risks, and suggest alternatives I may not have considered. Do not agree with me simply because I proposed the idea."

This small change can significantly alter the nature of an AI-assisted discussion.

The objective shifts from seeking approval to examining an idea more rigorously.

Of course, AI is not a human thinking partner, and asking it to be critical does not guarantee that its criticism will be accurate.

Its recommendations still require evaluation.

But the interaction can help users develop better questions and consider perspectives they might otherwise overlook.

The objective is not to make AI think instead of you. It is to use AI to strengthen your own thinking.

Why I Taught Participants to Question the First Answer

Another important lesson was the danger of treating an AI-generated response as automatically correct.

Many people assume that because AI produces structured, detailed, and confident answers, those answers must be reliable.

But generative AI can produce factual errors, outdated information, incomplete explanations, invented references, and unsupported conclusions.

Sometimes the first answer is accurate.

Other times, it may be partially correct or entirely misleading.

The real problem is accepting it without evaluation.

I encouraged participants to develop the habit of questioning important AI-generated information.

• Where did this information come from?
• Can I verify the sources?
• Does this explanation make sense?
• What assumptions has the system made?
• Could there be another interpretation?
• What information might be missing?

I also explained why asking AI to verify its own answer is not sufficient.

A system may repeat its original mistake or generate another convincing explanation.

Important claims should be checked against original research, authoritative documentation, reliable external sources, or appropriate expert knowledge.

This becomes especially important when using AI for academic work, engineering projects, business decisions, research, and professional recommendations.

AI literacy should include the ability to recognise when an answer needs further investigation.

From Passive AI Users to Independent Thinkers

One of my central objectives was to change how participants approached artificial intelligence.

I wanted learners to move beyond asking only what AI could generate for them.

I wanted them to begin asking what they could understand, question, evaluate, and build with its assistance.

That meant encouraging a shift from:

• Accepting answers to evaluating answers.
• Seeking agreement to inviting constructive criticism.
• Copying prompts to understanding how AI systems behave.
• Depending on AI to solve everything to developing independent problem-solving skills.
• Using AI simply for convenience to using it with purpose and responsibility.

This was the learning philosophy behind the masterclass.

The aim was not to turn every participant into an AI engineer or expert after one educational experience.

It was to provide a stronger foundation for further learning.

I wanted participants to leave understanding that AI literacy is not measured by the number of tools someone knows or the number of prompts they have memorised.

It is demonstrated through judgment, understanding, and responsible application.

Why I Believe Everyone Needs AI Literacy

Artificial intelligence is no longer relevant only to engineers, researchers, and technology companies.

It increasingly influences how people access information, communicate, study, work, and make decisions.

This creates a need for broader public understanding.

Students should know how to use AI to support learning without bypassing the reasoning process.

Professionals should understand how to evaluate AI-generated work and protect sensitive information.

Entrepreneurs should know how to question AI-assisted recommendations rather than simply accept them.

Parents and educators should understand the opportunities and limitations of AI as young people encounter it.

And everyday users should know that a confident AI response may still be incorrect.

These are not advanced technical requirements.

They are foundational skills for informed participation in an AI-enabled world.

I believe AI education should therefore help people develop the ability to think independently while taking advantage of technological progress.

Beyond Prompts Is More Than a Masterclass

AI Beyond Prompts was built around a simple conviction.

The future should not consist of people who know how to ask AI questions but struggle to evaluate the answers.

It should include people who can think critically, solve problems, examine evidence, and use technology with purpose.

Artificial intelligence will continue evolving.

New tools will emerge, capabilities will improve, and the way people interact with technology will change.

But the ability to reason independently, challenge assumptions, and make informed decisions will remain valuable.

I do not want people to become afraid of artificial intelligence.

I want them to understand it.

I do not want people to reject technological progress.

I want them to participate in it intelligently.

And I do not want AI education to stop at teaching prompts.

I want it to cultivate people who can question, evaluate, create, and solve meaningful problems.

Artificial intelligence should expand human intelligence, not become an excuse to stop exercising it.

That is the problem I wanted to address.

That is why I taught AI Beyond Prompts.

And that is why I believe learning how to think with artificial intelligence is becoming one of the most important digital skills of our time.

Research References

1. Lee, H.-P., et al. (2025). The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects From a Survey of Knowledge Workers. CHI 2025.

   Read the research

2. Zhai, C., Wibowo, S., & Li, L. D. (2024). The effects of over-reliance on AI dialogue systems on students' cognitive abilities: a systematic review. Smart Learning Environments.

   Read the research

3. Kosmyna, N., et al. (2025). Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task. Research preprint, arXiv:2506.08872.

   Read the research

Research note: These studies report associations, findings from specific experimental settings, and concerns about AI overreliance. They do not establish universal or permanent cognitive decline caused by artificial intelligence.

---

Ealeonore Mafor is an engineer and AI educator exploring artificial intelligence, engineering, digital strategy, and technology innovation.

Interested in learning to use AI beyond prompts?

Explore practical AI literacy education and training designed to help individuals and professionals use artificial intelligence with greater understanding, critical thinking, and responsibility.

Get in touch to learn more about AI literacy training and educational opportunities.