When AI can produce a fast answer, the value of education can't stop at getting the answer right. Students need to build the habits that help them question assumptions, weigh evidence, make ethical judgments, and create ideas that fit a new situation.
Kristina Kallas, Estonia's Minister of Education and Research, argues that higher-order thinking should become a central goal of education in the AI era. Her message is direct: schools should use AI in ways that make human thinking stronger.
AI forced Estonia to ask a different education question
When Kallas became Estonia's Minister of Education and Research in April 2023, ChatGPT had been publicly available for about five months. AI was not part of her original agenda, yet it quickly became hard to ignore.
An advisor raised the issue twice in the first months of her term. Estonia needed to do something about AI, he said, because students were already using it in school. However, neither of them had a ready-made policy response.
The real concern was human learning
Kallas shifted the question away from the technology itself. A ministry of education does not need to decide whether a tool is impressive. It needs to ask what that tool does to learning, attention, judgment, and the human ability to think.
To investigate, Estonia brought together neuroscientists, cognitive scientists, technology companies, and IT specialists. The group considered how AI might affect learning and what policy makers should do before schools drifted into using AI only for shortcuts.
Education leaders should focus less on the technology itself and more on what happens to human learning.
Computers raise the stakes for human skills
Kallas pointed to an OECD workforce skills survey from 2017. In her talk, she said that about two-thirds of workers at the time had literacy, numeracy, and digital skills below the level of computers then available.
That comparison creates an uncomfortable but useful question. If computers already outperform many people at calculating, retrieving information, and processing text, what should people spend their time learning?
Basic skills still matter, but they aren't the finish line
Reading, numeracy, and digital skills remain necessary. Without them, people can't properly assess what a system produces or spot an error in a confident-sounding answer. The OECD's discussion of a digital world of work also emphasizes strong cognitive skills, including literacy, numeracy, and basic ICT skills.
Yet schools can't treat remembering and repeating information as the end goal. AI can already handle much of that work quickly. Students need to know facts, but they also need to ask where facts came from, how they connect, and whether they apply to the problem in front of them.
AI creates cognitive pressure, much like the printing press did
Kallas compares this moment to the invention of the printing press. Reading was not once a universal skill. As printed material became more available, societies had a strong reason to teach more people to read.
AI creates a similar kind of pressure, although it is mainly cognitive rather than physical. The challenge is to help people think faster when needed, reason more systematically, and form original responses when familiar patterns no longer fit.
The brain prefers the easier route
Kallas describes two broad modes of thought. Lower cognitive processes rely on memory, understanding, routine, and repeated action. Higher cognitive processes require people to re-examine a situation, test their assumptions, and use knowledge in a new way.
Both modes are useful. The issue is that the brain naturally prefers the first one because it consumes less energy.
Automated skills save effort
Once you learn to ride a bike, ski, or skate, you do not start from zero every time. Your brain stores the basic skill, then allows you to perform it with little conscious effort. That efficiency frees attention for other things.
The same pattern appears in factual recall. If you remember that World War II began in 1939, you can state the date without rebuilding the history behind it. This is a useful ability, but it does not require fresh analysis.
The brain also uses shortcuts when it forms stereotypes. If someone absorbs the idea that Italians are loud, they may expect every Italian they meet to match that pattern instead of learning about the individual in front of them.
Familiar patterns are energy-efficient, but they can stop us from looking closely at a new person, problem, or piece of evidence.
Higher-order thinking takes real effort
Higher-order thought means learning and relearning. Knowledge that worked yesterday may not work in today's setting, so people need to adjust it rather than repeat it automatically.
This kind of mental work is tiring for a reason. The brain consumes a large share of the body's energy. Someone who spends long hours preparing for a difficult exam may feel physically drained afterward and need sleep to recover.
AI makes this distinction more important. If students let a tool take over every difficult mental step, they may get an answer while losing the practice that develops their own ability to analyze and adapt.
Education must move beyond remembering and applying
For roughly 200 years, many education systems have centered on a familiar sequence: learn material, remember it, understand it, apply it, and take a test. These are meaningful stages, but Kallas argues that they are no longer enough on their own.
Her cognitive "tree" separates lower-order and higher-order skills.
| Lower-order skills | Higher-order skills |
|---|---|
| Remembering information | Analyzing causes and relationships |
| Understanding what was learned | Evaluating evidence and ethical consequences |
| Applying a known answer or skill | Creating a response for a new situation |
The key shift is that schools need to help students climb past the apply stage much earlier.
Analysis asks why, not only what
A student can memorize that World War II began in 1939. Analysis asks why it began then, which actors shaped events, what motives drove their choices, and how political relationships changed the outcome.
That process requires a student to connect facts instead of treating them as isolated items. AI can help surface information, but the student still needs to judge which relationships matter and explain the reasoning.
Evaluation includes ethics and responsibility
Evaluation goes beyond checking whether an answer is factually correct. It asks whether an action is fair, responsible, justified, or harmful.
This matters when people use AI for school, work, hiring, marketing, public services, or creative projects. A tool can produce an option, but people must decide whether the option should be used and who may be affected by it.
Creation means more than generating output
Kallas distinguishes human creation from output generated through statistical probability. AI can produce drafts, images, and suggestions based on patterns in existing material. Human creation involves purpose, judgment, experience, and a response to the particulars of a new situation.
The goal is not to compete with AI at tasks it already performs well. The goal is to strengthen the human judgment required to use its output well.
Estonia wants students to practice deeper thinking earlier
Traditionally, students may encounter sustained analysis, evaluation, and original research later in university, especially in PhD programs. Kallas argues that this practice should begin far earlier.
In Estonia, lower secondary school begins around seventh grade, when students are often 12 or 13 years old. That is the point at which students should begin working more deliberately with analysis, evaluation, creation, critical thinking, and systematic reasoning.
Assignments need to demand more than a correct answer
Access to AI alone does not produce stronger thinking. The assignment design matters. If a teacher asks for a simple summary, students can easily use AI to produce one. If the teacher asks students to compare competing explanations, justify a position, identify missing evidence, and revise their conclusion, the student has to do more cognitive work.
Teachers become central to this approach. They need the confidence to use AI within a lesson while still setting clear expectations for thought, evidence, and original judgment.
For broader context on available platforms, a current AI education tools review shows how varied the market has become. Estonia's approach, however, focuses on the learning process rather than treating a general-purpose tool as the lesson itself.
Estonia's AI Leap program uses AI as a tutor
Estonia developed an AI strategy for schools by the end of 2024 and launched its AI Leap program in February 2025. The first pilot year began with students in grades 10 and 11 at upper secondary schools.
The program is planned to expand to vocational education students in September 2026. It operates as a public-private partnership, with half of the budget coming from private-sector companies and half from Estonia's Ministry of Education and Research.
A foundation runs the initiative
Rather than manage the program directly inside the ministry, Estonia assigned the work to a foundation. Kallas said ministries are not always the best place to run an innovative initiative that needs room for testing and adjustment.
The program's main priority is teacher competence. Teachers need to understand how AI can move students toward more analytical, systematic, creative, and critical work instead of allowing the tool to complete the work for them.
The tool is designed for Socratic tutoring
Estonia is working with a Socratic-style AI tutoring tool developed with OpenAI. Kallas stressed that it is not simply general-purpose ChatGPT placed in a classroom.
A Socratic tutor should push learners with questions. Rather than immediately handing over the answer, it can ask students to explain their reasoning, test a claim, consider another perspective, or use evidence to support a conclusion. This is how AI can become an assistant for higher-order learning.
Kallas discussed the program in her TEDxUniversity of Tartu talk, part of the independently organized TEDx event format.
"It is not about using AI the most. It is about how and why students use it."
The pilot is meant to produce evidence, not false certainty
Estonia does not claim to have settled every question about AI and learning. Kallas said the science is still limited, and the program is monitoring results while it operates.
That stance matters because schools face two bad options: wait indefinitely for perfect answers or adopt AI without studying what it does to students' learning habits. The pilot takes a third path by using the technology under a defined educational purpose and observing the results.
Avoiding cognitive offloading
Kallas named "brain rot" and cognitive offloading as risks Estonia hopes to avoid. Cognitive offloading happens when people hand too much thinking to a device instead of using their own memory, reasoning, and judgment.
A calculator can help with arithmetic, for example, but a student still needs number sense to recognize an impossible result. The same principle applies to generative AI. Students need enough understanding to challenge a weak answer, notice a false claim, and decide when a response lacks evidence.
The OECD Survey of Adult Skills overview offers useful context for why foundational skills remain important even as technology changes. A person cannot evaluate an AI-generated explanation without being able to read, reason, and question it.
Other countries can learn from the results
Kallas described Estonia as the first country to provide this type of Socratic study tool to students at this scale. Other education systems are watching, and Estonia intends to share what it learns.
The long-term value may be the evidence produced by the pilot: how AI affects the learning process, which teaching methods keep students mentally active, and where technology starts to replace thought instead of supporting it.
Human thinking remains the point of education
AI can retrieve, summarize, and generate at extraordinary speed. Those abilities make it even more important for schools to train students to analyze, evaluate, create, and relearn when circumstances change.
Estonia's experiment starts with a simple standard: AI use should have a reason. Success is not measured by how many hours students spend with a tool or how much work the tool completes.
The smartest use of AI may be the use that leaves people with stronger minds, better questions, and more responsibility for the answers they choose.
0 facebook:
Post a Comment