Short video cuts thought into fragments. AI may skip the process of forming an answer altogether. A stronger system does not necessarily mean a stronger user.
Part I examined how information feeds fragment attention and why information is not the same as intellectual capital. AI adds another layer. It need not make us switch constantly. It can present a complete, fluent and apparently credible answer as soon as a question appears.
That is an enormous capability. We no longer have to perform every search, organisation, translation and calculation alone, and scattered knowledge can be connected at speed. But the strength of the tool makes another risk easy to hide behind the result. The combined human-AI system performs better while the person’s own ability may not grow with it.
AI can make the system stronger without strengthening the person
With AI, reports are written faster, code appears faster, and an unfamiliar question quickly receives a well-structured explanation. Judged only by the final output, the person has clearly become more capable.
But at least two kinds of ability are involved. One is the ability to obtain a result quickly with a tool. The other is the ability to understand, produce and test that result independently. AI greatly expands the first. It does not guarantee an increase in the second.
Someone can receive an excellent answer without knowing its assumptions. They can submit a polished analysis but be unable to reconstruct the reasoning without the original text. They can ask AI to compare ten positions without developing a standard of their own for judging evidence. What has become stronger in that case is the human-AI system, not necessarily the human being.
The illusion is dangerous because the two states look almost identical in the short term. Whether AI has augmented a person or replaced them, fluent prose and a correct result may appear on the screen. The difference usually becomes visible only after the tool is removed.
An answer that arrives too quickly can end curiosity
Much extended thought begins in the unknown. We have no answer yet. The resulting tension keeps us looking for material, trying explanations, encountering contradictions and revising what we first believed.
More uncertainty is not always better. A problem with no point of entry merely makes people give up. Exploration is driven by a moderate unknown: I do not know the answer, but I can see part of its outline and believe that a few more steps may reveal something.
Research on curiosity and learning has found that curiosity can improve memory not only for an answer but also for other information encountered during the same period. ‘Wanting to know’ is not an incidental feeling that precedes learning. It changes how the brain receives information. See the research oncuriosity and memory mechanisms.
AI, however, tends by default to fill the unknown as quickly as possible. The question has barely been asked when a tightly organised answer appears. It may genuinely solve the problem, or merely make the problem look solved. Curiosity has been soothed before a model has formed inside the person.
This does not mean that slower answers are always better, or that efficiency should be rejected on principle. The order matters. If an answer defines the problem, decides which evidence matters and locates the conclusion before a person has built a structure for the question, that person usually receives a result rather than an act of thinking.
Is AI augmenting me or replacing me?
To decide whether AI augments or replaces a person, we should not ask only how much work it performed. We should ask what remains afterwards.
One simple test is this: after closing AI, can I reconstruct the logic in my own words? If an assumption changes, do I know why the conclusion changes with it? When I find a counterexample, can I revise the model rather than immediately asking AI for another answer?
A second test is to put the result in a new setting. Knowledge that has become intellectual capital can usually transfer. It does more than supply the answer to one problem. It lets a person see why another problem has the same structure. If the ability disappears as soon as the original prompt and wording are removed, the answer probably still belongs to the tool.
Empirical work in this area is developing rapidly, and the conclusions remain open. In one randomised experiment on generative AI in mathematics learning, an unrestricted general chat interface improved performance during practice but may have worsened performance on an independent exam after the AI was removed. A specially designed tutoring interface that restricted direct answers reduced the problem. The result suggests that the effect depends on whether AI supports practice or replaces it. See the research ongenerative AI and learning performance.
AI does not make people intelligent or unintelligent by nature. The long-term result depends on what it replaces and what it forces the person to practise.
Metacognition: first notice what is happening to you
The first form of resistance I can think of is metacognition. Put simply, it means stepping outside the behaviour for a moment and observing yourself.
Do I truly understand, or do I merely possess an answer? Why did I open AI at this moment? Am I removing mechanical work, or escaping a piece of thinking that I should have done myself? If I close the window, can I still explain the logic?
This does not happen automatically. The more smoothly AI answers, the more easily familiarity takes over: every sentence makes sense, so I assume I have mastered it. One way to break the illusion is to record my own judgement before asking, however rough it may be. After receiving the answer, I can close AI and try to reconstruct, derive and challenge it.
If I cannot reconstruct it, the answer is still inside AI. It has not entered my intellectual capital.
A multidimensional database: make AI a node, not the judge
The second approach is not to reject external tools but to place AI inside a larger cognitive structure.
In that structure, the human mind is the central node. Personal notes, primary material, lived experience and AIs with different functions form the other nodes. One AI retrieves. Another searches for counterexamples. Another makes connections across fields. Others can simulate different positions.
A useful database cannot save conclusions alone. It must retain sources, the strength of the evidence, degrees of uncertainty, conflicting explanations, and the conditions under which a view was formed. Otherwise, a growing database simply accumulates familiar claims and makes it harder to tell what deserves belief.
AI can be a powerful node, or even a cluster of nodes, but it should not be the final judge of the whole system. A person must still choose the questions, decide which evidence to accept and know when to revise a judgement. Only then does external intelligence extend the boundary of human cognition instead of covering human judgement.
Timeline thinking: do not be fooled by a complete answer at one moment
The third approach is to turn a static analysis into an observation made along a timeline.
A model may explain what has already happened beautifully without showing how events will change next. Timeline thinking keeps asking: how did this state form? Which variables changed first, and which appeared later? What outcome follows from the current judgement? Did earlier predictions come true?
A view preserved only in static prose can look correct forever. Once placed on a timeline, it must carry a date, conditions and testable expectations. Returning later gradually reveals what worked, what failed and which causal stories were explanations invented after the event.
A timeline also resists the shallowness AI can bring. A complete answer generated at one moment need not become the endpoint. It can remain a provisional version, waiting for later facts, counterexamples and new experience to test it.
Cognitive sovereignty cannot be outsourced
Metacognition, multidimensional databases and timeline thinking all lead to the same question. Once AI enters my cognitive system, who controls that system?
I can give AI the work of retrieval, organisation, calculation and even part of the reasoning. Human minds have always used paper, books and databases. There is no virtue in doing everything alone. But AI differs from earlier tools. It does not only store information. It organises evidence, explains causes and generates judgements on our behalf. Without noticing, a user can move from thinker to receiver of results.
We cannot decide whether AI has augmented someone by output speed and quality alone. We must also ask whether long use has left that person with more knowledge, models and methods of judgement they can call upon independently.
The dangerous state is one in which the system grows stronger while the person grows more dependent on it. Without AI, they cannot reconstruct the answer, identify its assumptions or notice where it may be wrong.
Real augmentation looks different. AI raises the processing power of the system, while collaboration with it teaches the person to ask better questions, seek stronger evidence, recognise counterexamples and revise models. Only then does AI extend a human being rather than replace one.
I do not object to obtaining answers from AI. I object when the answer arrives so quickly that the question has not formed inside me; when the expression is complete but I have not gone through the process of forming a judgement; and when the system has thought in my place but I mistake its ability for my own.
Some unknowns should therefore remain for a while. The confusion and discomfort they produce are not entirely inefficient. Sometimes an absent answer is precisely what keeps someone searching for material, finding contradictions, revising assumptions and slowly connecting things that seemed unrelated. Fill the unknown too early and thought may stop growing with it.
Cognitive sovereignty does not require me to reject AI. It requires me to retain the ability to ask questions, reconstruct answers, examine evidence and revise models after borrowing AI’s power.
I decide why a question deserves thought. I examine whether an answer deserves trust. I also bear the consequences of the judgement.
If those powers remain in my hands, AI is an extension of my cognition.
If I have handed them away, the prettier the answer on the screen, the more complete the illusion. Only the system has grown stronger. The person gradually losing the ability to ask questions, form judgements and bear their consequences is me.