common basis
↗Theoretical foundations
Brain, development, intelligence, individual differences, AI, screens, context and evidence.
Introduction
Cognitive
dependence.
Use AI without giving up your judgement.
Cognitive dependence is an informative website to understand how the relationship between thought and technology changes. It does not assume that AI is good or bad: it asks what task is delegated, what capacity needs to be preserved, who uses the tool and under what conditions.
Rigorous science communication, not a diagnosis. Every statement includes its scope, its limits and a supporting source.
Content structure
The common theory is applied after
Students
Families
Adults
Professionals
Organizations
Editorial principle No case study re-explains all the theory: it translates the foundations into decisions, risks and practices of each group.
Extensive common base
The theory is explained once, with depth and references.
No alarmism
Benefits, risks, associations and causality are separated.
Real differences
Age, skill, education, health and context modify the effect.
Concrete application
Each group receives its own decisions and practices.
Why does it exist?
The question is not how much you use AI.
It's what relationship you build.
People have always thought with the help of tools, symbols and other people. AI expands that possibility because it can write, explain, summarize, propose and decide at high speed. The same ease that frees up time can also hide what we have understood, what we know how to verify, or what we would be able to reconstruct without assistance. This project offers language and criteria to observe that difference.
Practical purpose
Help people use AI more intentionally and with less unintended dependence.
The website does not ask people to stop using AI or present the technology as a threat in itself. Its purpose is to turn a vague concern —‘perhaps I am delegating too much’— into observable questions: which goal remains human, which capability should be retained, which result must be verified and what would happen if the assistance disappeared. The intended outcome is not absolute self-sufficiency, but conscious, proportionate and reversible collaboration.
Distinguish what the person and the AI each think through, decide, produce or verify.
Retain the practice needed to understand, oversee, correct and respond to exceptions.
Align confidence with one's own knowledge, the available evidence and the consequences of error.
Benefit from assistance without making one particular tool the only possible route.
Understand
Explain how attention, memory, cognition, self-regulation, and judgment are formed throughout life.
Go to brain and learning →Distinguish
Differentiate help that expands capabilities from delegation that reduces the necessary practice, supervision or autonomy.
Understanding cognitive offloading →Apply
Translate evidence into realistic decisions based on age, training, profession, environment and task.
Choose a case study →Work concept
Cognitive dependence does not mean using a tool frequently.
It is a functional relationship: it appears when important performance is linked to help that the person can no longer supervise, replace or withdraw without a relevant loss of understanding, judgment or autonomy.
Prior knowledge, development, capabilities, motivation, health and confidence.
Objective, difficulty, novelty, consequences and need to learn it.
Reliability, help design, friction, explainability and verifiability.
Time, incentives, access, support, standards, language and alternatives available.
Aid can be withdrawn
The person maintains the objective, understands the essential, verifies the result and can respond to an exception.
The task requires the tool
It is not always negative: a calculator, a screen reader or a browser can be legitimate supports. The question is what happens if they fail.
What was still necessary was delegated
The person obtains the product, but loses the ability to detect errors, transfer learning or act autonomously.
Content architecture
First the common fundamentals.
Then, the case studies.
Age or profession do not organize scientific theory. They organize practical decisions. That is why the website explicitly separates the theoretical foundations from their application to each group.
common basis
Theoretical foundations
They build a common vocabulary and present theories, findings, controversies, and limits of evidence.
- Brain and learning
- Development throughout life
- Intelligence and constructs
- Individual differences
- AI and cognitive offloading
- Screens, social media and IQ
- Inequality and context
Contextual application
Case studies
They start from the same foundations, but select problems, decisions, examples and practices relevant to each reality.
- What should be protected
- What can be delegated
- How to adjust help
- Observable signs
- Judgment exercises
- Cases and recommendations
- Specific evidence
Editorial principles
What we commit to explaining.
And what we will avoid claiming.
The field changes quickly and mixes different disciplines. These rules serve to avoid turning a legitimate concern into a conclusion stronger than the data.
Neither technophilia nor rejection
Benefits and risks are assessed according to what AI is doing, not according to a prior stance on the technology.
Capacity is not performance
A better assisted product does not by itself demonstrate independent learning, transfer, or competence.
Correlation is not cause
For screens, social media, IQ and ageing, we distinguish study designs that support causal inference from those that do not.
Age is not destiny
The stages guide, but specific knowledge, health, context and individual differences can weigh more.
Models are not labels
Intelligence constructs are explained as different approaches, with different support and uses.
Access is not capacity
Socioeconomic, linguistic or accessibility barriers are not interpreted as lower cognitive potential.
Evidence method
References do not decorate the text: they delimit what can be said.
Each block distinguishes the type of study, the population, the outcome measured and the main limitation. When research on generative AI is still recent, it is combined with better-established cognitive mechanisms without pretending that both levels of evidence are equivalent.
See criteria and bibliography →Stronger causal inferenceRobust experiments and syntheses
They allow effects to be estimated under specific conditions, but do not guarantee that they generalize to any age, task or tool.
Consistent patternsLongitudinal studies and observational meta-analyses
They help track changes and associations, although they may maintain selection or confounding factors.
Early signalSurveys, cross-sectional studies and self-report
They help formulate questions and describe experiences; they cannot by themselves prove cognitive decline or improvement.
DebateHypotheses and extrapolations
They are presented as possibilities that require testing, never as established harms or benefits.
Case studies
Choose the reality you want to analyze.
The cases follow a stable order and do not replace common foundations. Each one selects the relevant mechanisms, translates them into real situations and incorporates their own recommendations and exercises.
Students
Transferable learning, prior knowledge, assessment and stage-adjusted help.
Families
Family guidance, use agreements, conversation, play, study and age-appropriate support.
Adults
Everyday use, lifelong learning, cognitive accommodation and active aging.
Professionals
Expert competence, responsibility, supervision and tasks that should not be delegated.
Organizations
Work design, governance, collective memory, training and operational resilience.
Starting point
Before deciding how to use AI, it's important to understand what we still want to be able to do.Start with the theoretical foundations →