Cognitive augmentation or atrophy: how to use AI without losing your critical judgment
AI saves you time. It summarizes, organizes, drafts, and suggests solutions in seconds. But if you also hand over the thinking, the opposite of what you want can happen: over time, you find it harder to analyze, remember, and make decisions on your own.
This chapter begins the second module in the series, focused on the relationship between AI and the human mind. You will see why AI differs from other technologies that have helped us think, what kinds of mental effort are worth delegating, and which are worth keeping. You will also find three practical protocols for using AI as support without weakening your professional judgment.
The paradox of delegating mental effort
For centuries, people have been transferring mental functions to external tools. This is known as cognitive offloading.
- Writing allowed us to store memories and knowledge outside our memory.
- The printing press made knowledge accessible to many more people.
- The calculator took over basic arithmetic operations.
In all these cases, delegating one task freed up mental space for more valuable ones: analyzing, connecting ideas, and creating.
What changes with generative AI
Generative AI is different. It does more than store information or perform mechanical calculations. It can take over entire stages of reasoning itself: structuring an argument, drawing conclusions, or developing a strategy.
If you routinely use it passively to solve problems that previously required effort from you, two effects can emerge:
- The illusion of competence: AI gives you a clear, quick summary, which makes you feel you have mastered the subject. In reality, you have not worked through it. That feeling reduces your desire to study or check what you read.
- The loss of skills (deskilling): the brain primarily maintains the abilities it uses. If you stop practicing systematically, your working memory, ability to follow long lines of reasoning, and tolerance for ambiguity weaken.
That is why the challenge of AI is not just technical. It is also mental. The key question is this: how can you use AI as an exoskeleton that makes you stronger, rather than as a wheelchair that prevents you from walking?
Which mental effort to delegate and which to keep
Cognitive load theory, developed by psychologist John Sweller in the late eighties, is very helpful in answering that question. This theory distinguishes three types of mental effort. What matters is not just how much effort a task requires, but what that effort is spent on.
Where AI should step in: extraneous load
AI is very useful for eliminating mechanical work that adds nothing to your learning or decisions. For example:
- Cleaning, transforming, and organizing data.
- Performing an initial search and filtering pass in large document repositories.
- Producing routine drafts from templates you have already defined.
Where it should not replace you: germane load
The risk arises when you delegate to AI the very effort that makes you a better professional:
- Defining what the problem really is.
- Weighing the ethical aspects of a decision.
- Formulating hypotheses without having reflected first.
If the machine does the work of testing ideas against one another and building an argument from scratch for you, no new schemas form in your long-term memory. Over time, your expert judgment suffers.
AI as scaffolding, not a prosthesis
In educational psychology, scaffolding is temporary support that helps someone do what they cannot yet do alone. The concept comes from the work of David Wood, Jerome Bruner, and Gail Ross. It is related to Lev Vygotsky's zone of proximal development, which is the distance between what you can do on your own and what you can do with help.
Applied to AI, the idea is for the system to support you while you think, not to think for you.
The wrong approach: cognitive substitution
Example prompt:
“Analyze this case and tell me the best strategy.”
Here, all you do is receive conclusions.
The scaffolding approach: cognitive augmentation
Example prompt:
“I have reached preliminary conclusion X based on factors A and B. Act as a critical auditor: identify three methodological biases in my approach and propose two plausible alternative explanations.”
In this case, you do the thinking, and AI acts as an adversary that forces you to refine your reasoning.
Three protocols to protect your professional judgment
These three habits help you stay in control of your own thinking in your daily work.
Protocol 1: think first, ask second
Before asking AI to solve a complex problem, write down your own hypothesis or, at a minimum, the criteria you expect a good answer to meet.
There is a simple rule: if you do not know how to evaluate the answer before receiving it, you will not be able to review it properly once you have it either.
Example: before requesting a contract risk analysis, write down the three risks you believe exist. Then compare your list with AI's and analyze the differences.
Protocol 2: add friction deliberately
For important decisions, do not ask for a direct answer. Configure your agents to proceed step by step and make you think before giving you a draft:
- Have them ask you questions to clarify the case.
- Have them ask you for the missing data.
- Have them present you with dilemmas or alternative scenarios.
This friction activates what psychologist Daniel Kahneman called System 2: slow, reflective, analytical thinking, as opposed to quick, intuitive responses.
Example system instruction:
“Before making any recommendation, ask me at least three questions about the missing information and wait for my answers.”
Protocol 3: audit a sample
Periodically, randomly select some analyses or reports generated by your AI systems and review them thoroughly, as if you were starting from scratch:
- Check every original source.
- Recalculate the key figures.
- Ask yourself whether the conclusion holds up without the accompanying text.
This exercise, similar to red-teaming (attacking your own system to find its flaws), has two benefits: it detects errors that would have gone unnoticed and keeps your own analytical abilities sharp.
Conclusion
AI can expand your abilities or weaken them. It depends on how you use it. Delegate mechanical, repetitive work to it, but keep the effort that helps you think better: forming hypotheses, testing ideas against one another, and making decisions. If you use AI as scaffolding, you gain speed without losing the most valuable part of your work: your professional judgment.
Want to build AI systems that help you think better, not less? Visit libertadia.com and sign up in the app.