Human Intelligence: Origins, Cognitive Mechanisms, Development and the New Frontiers of Artificial Intelligence

Intelligence remains one of the most complex phenomena to define, measure and understand. Long associated with the ability to reason, master language and produce knowledge, it is now understood through a plurality of dimensions: adaptation to the environment, problem-solving, learning, mental representation, abstraction, memory, creativity and the ability to establish connections between apparently unrelated phenomena.

This diversity raises a fundamental question: is intelligence a general capacity whose manifestations vary between individuals, or does it result from the organisation of multiple skills developed through biological evolution and learning?

Studying intelligence requires more than observing intellectual performance. It also involves examining the conditions under which cognitive abilities emerge, the transformations that progressively organise them, and their relationships with action, language, emotions and social life.

Historical developments in experimental psychology have provided several answers to these questions. Alfred Binet’s research into the measurement of intellectual abilities, Charles Spearman’s analysis of the general intelligence factor, Jean Piaget’s investigations into the development of mental operations, and Henri Wallon’s work on the relationships between intellectual development, motor activity and emotions have all helped shape the field.

These approaches, however, do not offer interchangeable explanations. Some focus on measuring individual differences, while others investigate the mechanisms of reasoning or the origins of mental representations. Examining their contributions together helps explain why intelligence cannot be reduced to a single indicator or equated with the mere accumulation of knowledge.

This distinction has taken on renewed importance with the development of artificial intelligence. Contemporary systems can generate texts, solve exercises, write code, summarise documents and identify relationships between information drawn from different fields. These capabilities require a more precise definition of the criteria used to characterise intelligent activity, without confusing functional effectiveness with understanding, lived experience or cognitive autonomy.

To understand intelligence empirically, several levels of analysis must therefore be distinguished: observable abilities, the operations that make them possible, their development throughout life, and the biological, emotional, educational and social conditions that contribute to their organisation.

1. Defining Intelligence: A General Ability or a Collection of Functions?

1.1. From the Faculty of Knowing to Problem-Solving

In the classical conception, intelligence is understood as a faculty of knowledge for which language serves as a privileged instrument. This definition emphasises a characteristic feature of human thought: the ability to transform experience into representations and then manipulate those representations to produce reasoning.

Intelligence, however, cannot be reduced to this discursive activity. Behaviour observed in different animal species demonstrates capacities for adaptation, learning and problem-solving that do not depend on human language.

When confronted with an obstacle, an animal may modify its behaviour, explore several possibilities, use an object as a tool or exploit a spatial relationship to achieve a goal. Such behaviour does not necessarily amount to a simple reflex response. To varying degrees, it involves taking account of the properties of the environment and reorganising the means available.

Broadening the definition of intelligence therefore means considering it as the capacity to process a situation, identify relevant relationships and adjust behaviour according to a desired outcome.

This functional definition has an important advantage: it makes it possible to compare organisms or systems that do not use the same mechanisms. It also has a limitation. Effective behaviour, on its own, does not establish that the system concerned understands the situation in the same way as another system.

Successful performance, the nature of the operations involved and the degree to which a situation is represented internally are therefore three distinct objects of analysis.

1.2. General Intelligence and Specialised Abilities

One of the major questions in differential psychology concerns the existence of a general intellectual capacity that contributes to success across a variety of tasks.

At the beginning of the twentieth century, Alfred Binet developed an assessment method intended to evaluate certain intellectual abilities in children, particularly in the context of schooling. His approach helped establish the possibility of psychological measurement based on standardised tests.

This approach does not mean that intelligence can be reduced to a number. Its purpose is to observe performance under defined conditions and interpret the results within an assessment framework.

Charles Spearman developed this line of research by studying correlations between scores obtained on different tests. He proposed distinguishing a general factor, designated by the letter g, from specific factors associated with the particular demands of individual tasks.

In this model, the general factor accounts for part of the shared variation between different cognitive performances. It is associated with the ability to perceive relationships, identify patterns and organise elements into coherent structures.

Specific factors, meanwhile, correspond to the demands of particular activities, including verbal reasoning, spatial manipulation, numerical processing and the resolution of specialised problems.

Other approaches favour a more differentiated organisation of abilities. Between the idea of a completely general intelligence and that of a collection of independent skills, several models propose intermediate factors shared by groups of tasks.

These disagreements reveal an essential methodological difficulty: the statistical structure of performance does not, by itself, provide a complete explanation of the mechanisms that produce it.

A correlation between results may reveal a common underlying organisation, but it does not automatically establish the origin of that organisation, the cognitive processes involved or the conditions under which it develops.

1.3. Measuring Performance Is Not the Same as Explaining Its Origins

Intelligence tests make it possible to compare performance obtained under defined conditions. Their interpretation nevertheless depends on the nature of the tests, the reference population, the testing conditions and the abilities actually being assessed.

For example, an individual may demonstrate considerable verbal fluency without displaying the same ease in a spatial task. Another person may solve a practical problem efficiently while struggling to explain the reasoning behind the solution.

These differences do not necessarily make measurement useless. They make it essential to specify what is being measured.

Psychometric measurement primarily addresses the following question: how are performances distributed, and to what extent do they vary together?

Cognitive analysis seeks instead to determine which operations are involved in successfully completing a task.

The developmental approach examines how these operations emerge, change and become organised throughout life.

Finally, contextual analysis investigates the role of learning, experience, cultural environments and opportunities to practise particular skills.

These levels of analysis complement one another, but they should not be confused. Statistical measurement describes a phenomenon; cognitive theory proposes an explanatory mechanism; developmental research seeks to reconstruct the process through which that mechanism emerges.

1.4. Willpower, Interest and Perseverance in Intellectual Activity

Intellectual activity does not depend exclusively on information-processing abilities. It is also influenced by motivation, interest, engagement and perseverance.

An individual may possess the resources necessary to solve a problem but still fail if they give up too early, interpret difficulty as a threat or see no reason to continue.

Conversely, strong motivation may encourage someone to explore several solutions, seek additional information and revisit their reasoning after an unsuccessful attempt.

These factors should not automatically be equated with intelligence itself. Nevertheless, they influence how intellectual abilities are mobilised and the results they produce.

This distinction is particularly important in learning, scientific research and creative work. Success in these fields depends not only on access to knowledge but also on the ability to sustain an approach, revise a hypothesis and tolerate periods of uncertainty.

2. Practical and Discursive Intelligence: Two Ways of Processing Reality

2.1. Solving a Problem by Acting on It

Practical intelligence manifests itself when an organism adapts its movements and behaviour to transform a concrete situation.

An obstacle prevents access to a resource. A detour may be needed, an object may have to be moved, or an intermediate tool may be required to reach the desired object.

This problem-solving process may proceed through successive trials. The organism experiments with different behaviours, retains some effective responses and gradually abandons unsuccessful ones.

A solution may also emerge more rapidly when the arrangement of elements makes it possible to perceive a previously unnoticed relationship. This phenomenon is often referred to as insight.

In such cases, a solution appears when the situation is mentally or perceptually reorganised. The elements themselves may not have changed; rather, their functional relationship has become apparent.

Observations of animal behaviour have helped demonstrate that problem-solving does not always require a prior verbal formulation.

It is therefore possible to recognise complex adaptive abilities without attributing to animals the same forms of symbolic reasoning that characterise human societies.

2.2. Discursive Intelligence: Pausing Action in Order to Think

Discursive intelligence introduces another possibility: representing a situation before intervening in it.

Instead of responding solely to immediate constraints, an individual can examine several scenarios, compare solutions and anticipate their consequences.

Language plays a decisive role in this process. It makes it possible to distinguish the elements of a problem, name their properties, formulate hypotheses and preserve a record of reasoning.

A situation can thus be broken down into several questions: which facts are established? Which relationships appear plausible? What causes might explain the phenomenon? What observations would help distinguish between the hypotheses?

This activity requires differentiation and recombination. Elements initially perceived as a whole are isolated, examined and then placed within a more general structure.

Discursive thought also makes it possible to move partly beyond the constraints of the present. An individual can reflect on a past event, imagine a situation that has never occurred or construct a scenario concerning an uncertain future.

Language therefore greatly extends the scope of intellectual activity: mental processes no longer depend solely on what is directly accessible in the surrounding environment.

2.3. Representing the World: The Role of Symbolic Function

Representation involves mentally evoking an object, action or situation in its absence.

A word can refer to an absent object. A map can represent a territory. A mathematical formula can express a quantitative relationship without requiring the physical manipulation of the elements concerned.

Symbolic function makes it possible to replace directly perceptible reality with a sign, an image or a mental construction.

Henri Wallon emphasised this capacity for representation and the role of symbolic function in the development of human intelligence.

The sign is not the object itself. It refers to the object through conventions or learned relationships. Once this distinction has been mastered, representations can be manipulated, combined and used to construct reasoning that does not require the physical presence of the objects concerned.

This operation profoundly transforms the possibilities of thought.

A traffic problem can be studied using a map. An astronomical phenomenon can be described through equations. An economic theory can be examined by constructing models. In each case, a representational system makes it possible to investigate relationships that would be difficult to grasp directly.

Symbolic function is therefore one of the foundations of cumulative knowledge. Representations can be communicated, preserved, discussed and transformed by several individuals over time.

2.4. Continuity Between Action and Representation

The distinction between practical and discursive intelligence does not mean that these two forms operate independently.

From a perspective inspired by Jean Piaget, representation can be understood as an internalised extension of action. Mental operations develop progressively through interactions between the organism and its environment.

Other approaches, particularly those associated with Henri Wallon, place greater emphasis on the specific nature of symbolic function and the transformations it introduces into mental activity.

These approaches highlight different aspects of cognitive development. One stresses continuity between action and thought; the other draws attention to the change in cognitive organisation associated with the emergence of symbolic representation.

This difference should not be turned into an absolute opposition. Human intellectual abilities depend on interactions between perception, movement, memory, language and reasoning.

Abstract thought does not eliminate concrete experience. Rather, it makes it possible to reorganise that experience, compare it with other experiences and integrate it into more general systems of relationships.

3. The Development of Intelligence in Children: From Action to Conceptual Thought

Intellectual development does not consist simply of accumulating information. It involves a progressive transformation of the operations through which children perceive, represent and organise their environment.

The stages traditionally described in developmental psychology provide a useful framework for understanding this evolution. They should not, however, be interpreted as a rigid universal timetable: abilities vary between individuals, across domains and according to learning conditions.

3.1. From Birth to Three Years: Practical and Symbolic Foundations

During the first years of life, intellectual abilities develop in close association with motor maturation, social interaction and the evolution of communication.

Children progressively learn to coordinate their perceptions and actions. They recognise patterns, anticipate certain consequences and adjust their behaviour to the responses of those around them.

An Initially Social Form of Intelligence

Affective exchanges provide an essential foundation for early learning.

Facial expressions, smiles, vocalisations and gestures contribute to the establishment of relationships with adults. Children gradually learn to distinguish certain attitudes, solicit assistance and adjust their behaviour according to the responses they receive.

These behaviours are not merely communication in the narrow sense. They also make it possible to act upon the social environment.

The ability to interpret other people’s responses and produce understandable signals therefore contributes to the organisation of intellectual activity.

Manipulation and Concrete Problem-Solving

As movement and grasping abilities develop, children explore the properties of objects more directly.

They may move around an obstacle, displace an object or use a rudimentary tool to reach something beyond their immediate grasp.

Such behaviour reflects the progressive organisation of relationships between means and outcomes.

Practical intelligence then relies on the coordination of movement, perception and environmental properties. Success depends on exploiting a concrete configuration rather than explicitly formulating a general rule.

Deferred Imitation: A Decisive Transition

Imitation enables children to reproduce behaviour observed in others. When imitation occurs after the model has disappeared, it suggests that the action has been retained in a representational form.

Deferred imitation is therefore an important indicator of developing representational abilities.

Children no longer need to reproduce what they observe immediately. They can retain a trace of the action and reproduce it later.

Behaviour begins to become detached from the context in which it was learned.

Symbolic Play and the Emergence of Substitution

Pretend play offers a particularly clear illustration of symbolic function.

A child may pretend to sleep, turn a piece of paper into an imaginary sweet or attribute human intentions to a doll.

An object or action then acquires a meaning that goes beyond its immediate material properties.

The simulated gesture is no longer identical to the real gesture. The object being used is no longer identical to the object it represents.

This capacity for substitution opens a space in which representations can be combined, transformed and shared.

Language develops alongside this process. Words make it possible to identify objects, distinguish actions and progressively organise experiences into more stable categories.

3.2. From Three to Six Years: Stabilising, Ordering and Connecting Representations

During the preschool years, representational abilities progress rapidly, but their organisation remains variable across situations.

Children may recognise an object, talk about it and use it in symbolic play without mastering all the logical relationships required to compare it with other objects.

Several characteristic forms of activity progressively help structure their thinking.

Stabilisation: Identifying Objects and Their Properties

To develop concepts, children must learn to recognise certain characteristics as sufficiently stable despite changes in circumstances.

They gradually learn to distinguish objects, name them and preserve their identity across different experiences.

This stabilisation is important because organised thought cannot function effectively if objects continually change meaning according to immediate perception.

Early representations nevertheless remain partial. Children may focus on a particularly visible characteristic while overlooking other important properties.

A drawing, for example, may reproduce what a child knows about an object rather than its exact appearance from a particular viewpoint.

This way of representing an object does not necessarily indicate a simple failure of observation. It also reveals the role of existing knowledge in perception and representation.

Ordering: Organising Experience in Time

Narrative is a major intellectual activity.

When recounting an experience or telling a story, children must transform a set of representations into a sequence of propositions.

This operation involves selecting elements, determining their order and constructing temporal relationships.

Early narratives may juxtapose events without clearly specifying how they are connected. Connectors such as “and”, “then” and “afterwards” help the story progress, but they do not always express causality.

A sequence of events does not automatically constitute an explanation.

Learning to tell a story therefore involves distinguishing what happens first, what follows and what actually causes an event.

This development contributes to discursive thought, which depends on organising representations within a mental and temporal framework.

Coupling: Searching for Relationships

Children also begin to understand the world by linking phenomena in pairs.

They may associate rain with wind, smoke with the sky or night with sleep. These connections reveal a desire to explain, even when the proposed relationships remain imprecise.

Coupling represents an important stage in the development of relational thought.

It makes it possible to move beyond merely juxtaposing objects and begin searching for connections between them.

The difficulty lies in determining whether an established relationship is relevant, whether it applies across several situations and whether it corresponds to a genuine causal dependency.

A frequent association is not necessarily an explanation. This distinction becomes fundamental to the development of scientific reasoning.

Syncretism: Perceiving a Whole Before Distinguishing Its Parts

Syncretism refers to a form of understanding in which perceptions, actions, desires and explanations remain closely intertwined.

Children may perceive a situation as a coherent whole without being able to isolate its elements or specify their relationships.

This organisation helps explain certain difficulties in distinguishing a phenomenon’s appearance, its cause and the personal interpretations it evokes.

It can also encourage teleological or animistic explanations in which events appear to occur according to an intention or purpose.

These forms of thought must be understood within their developmental context. They reflect an effort to make the world intelligible using the representational resources available at a particular stage.

Intellectual progress does not consist of abandoning every global interpretation. It consists of becoming able to distinguish the dimensions of a situation, examine their relationships and test explanations against the facts.

3.3. From Six to Eleven Years: Analysis, Comparison and Logical Operations

During the school years, the development of analysis and comparison makes it possible to organise representations according to increasingly explicit criteria.

Children learn to isolate certain properties, compare objects and construct categories that no longer depend exclusively on immediate appearance.

This development facilitates the acquisition of intellectual skills used in reading, writing, arithmetic and scientific activities.

Analysis: Distinguishing Properties

Analysis consists of breaking down a situation or object in order to examine its characteristics.

An object can be studied in terms of its shape, size, colour, weight or function. These properties can then be compared independently of one another.

The difficulty lies in maintaining one criterion while temporarily disregarding the others.

A larger object is not necessarily heavier. A longer distance does not always mean a longer duration. A change in shape does not automatically imply a change in quantity.

Concept formation therefore requires the progressive differentiation of the dimensions of experience.

Comparison: Establishing Relationships Between Objects

Comparison becomes more rigorous when relevant properties can be isolated.

To arrange three sticks according to length, a child must recognise the same dimension in each one and determine their respective relationships.

The middle stick is longer than the first and shorter than the last. The relationship is no longer based solely on a contrast between two objects; it forms part of an ordered series.

This operation opens the way to transitivity.

If A is smaller than B, and B is smaller than C, then A is smaller than C.

The ability to infer a relationship that has not been directly observed represents a major advance in reasoning.

Seriation: Ordering Differences

Seriation consists of arranging elements according to the degrees of a shared property.

It may involve length, size, weight or any other dimension that can be ordered.

This operation requires the construction of a coherent system of relationships. Each element occupies a position that can be defined in relation to the others.

Seriation also facilitates quantitative operations and the understanding of measurement, since measurement relies on comparing quantities according to a common scale.

Classification: Creating Categories

Classification is based on identifying a property shared by several objects.

The same element can belong to different categories depending on the criterion used. An animal, for example, can be classified by species, habitat or diet.

The ability to change criteria makes it possible to construct hierarchical classifications and nested categories.

This ability is essential to conceptual thought. It makes it possible to identify shared properties among different objects and define the boundaries of a category.

Conservation: Distinguishing a Property from Its Appearance

Jean Piaget’s research examined how children understand the conservation of certain quantities despite perceptual transformations.

A quantity of liquid can be poured into a narrower container. The liquid level changes, but the quantity does not necessarily change.

To solve this type of problem, children must learn to distinguish a change in appearance from an actual transformation of quantity.

The conservation of substance, weight, volume and length does not necessarily develop at the same pace.

Conservation tasks illustrate a general question: under what conditions can a property be considered invariant despite observed transformations?

This question is also central to the physical sciences, where understanding a phenomenon often requires distinguishing quantities that vary from those that remain constant.

3.4. From Around Eleven Onwards: Formal Reasoning and Hypothetical Thought

The progressive development of formal reasoning makes it possible to manipulate propositions without systematically referring to objects that are physically present.

Adolescents can examine a hypothesis, derive consequences from it and compare several possibilities before deciding which is best supported.

This development transforms the conditions of reasoning.

It becomes possible to reason from statements whose truth has not yet been established, in order to determine what would follow if they were provisionally accepted.

Hypothetico-Deductive Reasoning

A hypothesis is a proposition examined as the starting point of a line of reasoning.

Deduction consists of determining the consequences that follow from that proposition and the rules being used.

This ability does not mean that the initial hypothesis is true. It makes it possible to distinguish the logical consistency of an argument from the truth of its premises.

In science, this distinction is essential. An argument can be logically valid while relying on an empirically false hypothesis.

Intellectual activity must therefore combine logical analysis with the examination of observations and data.

Combinatorial Thinking

Combinatorial thinking involves examining several factors and considering the different configurations they can produce.

In an experiment, it may be necessary to determine whether a phenomenon depends on temperature, the concentration of a substance, exposure time or an interaction between several variables.

To identify relevant relationships, it is important to avoid changing too many factors simultaneously without being able to distinguish their effects.

Combinatorial thinking makes it possible to construct a set of hypotheses, identify the possibilities to be tested and organise observations.

It is one of the foundations of experimental reasoning.

Historical Thinking: Understanding Transformations

Conceptual intelligence is not limited to logical abstraction. It also involves the ability to reconstruct how phenomena develop over time.

Understanding an object or institution may require examining its origins, transformations and conditions of evolution.

This historical approach involves two complementary directions: looking back to understand how a situation emerged, and considering possible future developments in light of the available knowledge.

Projection into the future does not amount to certain prediction. It relies on hypotheses whose validity depends on the quality of information and the stability of the relationships being studied.

The distinction between retrospective explanation and anticipation is particularly important when dealing with complex phenomena such as the evolution of societies, technologies and ecosystems.

4. Intelligence, Emotion and Personality: An Interconnected Organisation

4.1. Thought Does Not Develop in an Emotional Vacuum

Intelligence and emotion have often been studied as separate dimensions of personality.

Yet intellectual activity always takes place within a lived situation. It depends on the attention given to a problem, the interest it generates, the feeling of safety or threat, and the ability to sustain effort.

Henri Wallon placed particular emphasis on the relationships between emotional development, motor activity and intellectual activity.

From this perspective, attitudes constitute an essential level of organisation. They shape the individual’s relationships with themselves and with the external world.

Emotional attitudes contribute to the development of self-awareness and the organisation of social relationships. Intellectual attitudes make it possible to distinguish objects, organise knowledge and structure representations of the world.

These dimensions influence one another.

Emotional states can facilitate or disrupt concentration. Understanding a situation can change the emotions it evokes. Experiences of success or failure can transform expectations and subsequent behaviour.

4.2. Motivation and Perseverance in Problem-Solving

A difficult problem can be abandoned, explored methodically or approached through a series of attempts.

These differences depend partly on available knowledge, but also on engagement with the task.

Perseverance makes it possible to continue searching despite the absence of an immediate solution. Interest helps sustain attention. The ability to tolerate uncertainty allows a conclusion to be postponed when the available evidence remains insufficient.

These dispositions influence observable results without being identical to cognitive skills.

It is therefore useful to distinguish three dimensions: the ability to perform an operation, the willingness to use that ability, and the conditions that allow it to be sustained over time.

This distinction applies to classroom learning as well as scientific research, artistic creation and professional activity.

4.3. Emotion as a Factor in the Direction of Knowledge

Emotion is not merely a potential obstacle to reasoning. It also helps direct attention and determine which questions appear important.

Curiosity encourages exploration. Surprise can lead someone to question a familiar explanation. Dissatisfaction with an incomplete answer can motivate a search for more precise solutions.

However, emotions and motivations can also encourage biased interpretations.

Someone attached to a hypothesis may give greater weight to evidence that supports it and overlook contradictory information. A need for certainty may lead to the premature adoption of an explanation.

Intellectual development therefore requires not the elimination of all emotional involvement, but the ability to identify the conditions under which it influences judgement.

Critical thinking notably requires distinguishing personal conviction, logical consistency and the empirical strength of a proposition.

5. From the Psychology of Intelligence to the Scientific Understanding of Cognitive Abilities

5.1. The Limitations of a Static View of Intelligence

Models based on performance measurement have made it possible to identify important regularities.

Factor analysis identifies dimensions shared by several tests and makes it possible to study the structure of individual differences.

However, a statistical structure does not suffice to reconstruct the development of abilities.

Two individuals may obtain comparable results while using different strategies. The same person may also solve a problem in several ways depending on experience and the constraints of the situation.

Studying the origins of cognitive operations therefore provides an additional perspective.

It seeks to understand how initially limited abilities develop into organised systems of representation, comparison and reasoning.

This perspective goes beyond a conception in which intelligence consists simply of a collection of juxtaposed factors.

5.2. Intellectual Operations and Their Organisation

Cognitive abilities do not necessarily function as independent units.

Analysis prepares the ground for comparison by making it possible to isolate relevant properties. Comparison facilitates classification. Seriation organises differences according to an ordered relationship. Logical operations then build on these achievements to construct more complex forms of reasoning.

This progressive organisation shows that some operations create the conditions necessary for others.

Mastering a skill therefore depends not only on possessing an elementary ability. It may also depend on coordinating several operations.

Mathematical reasoning, for example, involves knowledge, symbolic representations, logical rules and the ability to monitor the reasoning process.

A difficulty in one of these elements can affect overall success, even when the other abilities are relatively well developed.

5.3. Biological Maturation and Learning Conditions

Cognitive development is linked to the maturation of the nervous system, but it cannot be explained by maturation alone.

Experience, social interaction, education and opportunities to practise particular skills all contribute to the organisation of intellectual abilities.

An ability may remain underdeveloped when opportunities to use it are rare. Conversely, regular practice can improve performance in a specific domain.

It is important, however, not to conclude that training in one area necessarily produces a general improvement in all intellectual abilities.

The effects depend notably on the nature of the task, the duration of learning and the degree of similarity between the skills being trained and those subsequently assessed.

The distinction between maturation and learning does not mean that one must be chosen over the other. It helps explain that development results from complex interactions between the characteristics of the organism and the conditions of its experience.

5.4. Intellectual Productions as Evidence of Cognitive Diversity

Intellectual abilities can manifest themselves in very different productions: scientific research, literature, philosophy, engineering, music, visual arts and practical problem-solving.

These domains do not all require the same combinations of knowledge and operations.

Scientific research notably involves formulating hypotheses, analysing data and comparing models with observations.

Literary creation draws on command of language, the construction of meaning, narrative organisation and an understanding of a language’s expressive possibilities.

Technical activity often requires an understanding of material constraints, the ability to anticipate consequences and the adaptation of solutions to real-world conditions.

These differences do not establish a general hierarchy between forms of intelligence.

Rather, they show that performance results from particular combinations of skills, experience and attitudes.

The diversity of intellectual production therefore supports a multidimensional analysis of intelligence, attentive both to shared mechanisms and to specific forms of organisation.

6. Language: An Instrument of Thought, a System of Representation and a Means of Transmission

6.1. Naming in Order to Differentiate

Language makes it possible to stabilise distinctions and make them available to thought.

Naming a property helps distinguish it from other properties. Having words for cause, consequence, possibility or contradiction facilitates the organisation of reasoning.

This function does not mean that all thought depends exclusively on words. Visual, spatial, motor and emotional representations also play important roles.

Language nevertheless provides a particularly powerful system for making relationships between elements of a situation explicit.

It allows an experience to be transformed into a proposition and that proposition to be compared with others.

6.2. Inner Speech and the Control of Mental Activity

Some language activity does not occur aloud.

Inner speech can accompany the preparation of an action, the examination of a problem or the checking of a solution.

An individual can give themselves instructions, reformulate a question or consider the consequences of a decision before acting.

This activity helps organise reasoning and keep certain elements of a problem available during processing.

It also makes it easier to revisit a proposition, modify a plan and examine several interpretations.

Language thus becomes an instrument for controlling intellectual activity as well as a means of communication.

6.3. The Cumulative Transmission of Knowledge

Human representations can be preserved in narratives, texts, diagrams, archives and systems of notation.

This preservation makes it possible to transmit knowledge beyond individual experience.

Successive generations can revisit earlier findings, criticise them, improve them or integrate them into new theories.

Scientific knowledge illustrates this cumulative property particularly clearly. A discovery does not depend solely on the abilities of the researcher who produces it; it forms part of a body of methods, data and concepts developed collectively.

Human intelligence therefore has a social and historical dimension. Individual abilities are exercised within cultural environments that make symbolic tools and established knowledge available.

The possibility of transmitting representations greatly extends the scope of thought by providing access to experiences and lines of reasoning that no individual could reconstruct alone.

7. Artificial Intelligence: From Observable Performance to the Question of Understanding

7.1. A Functional Definition of Artificial Intelligence

Artificial intelligence encompasses computational methods that enable systems to perform tasks associated with certain cognitive abilities, including classification, prediction, content generation, searching for solutions, pattern recognition and language processing.

These systems do not all rely on the same architectures or learning principles.

Some use explicit rules. Others learn patterns from examples. Contemporary generative systems can produce new sequences of text, images, code and other content in response to a prompt.

Their behaviour may be described as intelligent in a functional sense when their performance demonstrates an ability to process a complex situation or produce a relevant solution.

This classification does not, however, automatically establish understanding comparable to that of a human being.

Observable results, the computational mechanisms that produce them and any mental properties attributed to the system must be distinguished.

7.2. Language Models and the Representation of Relationships

Large language models learn from extensive collections of textual data and, depending on the system, other types of content.

During training, they adjust their parameters to produce outputs suited to the patterns encountered and the learning objectives used.

Their operation enables them to generate structured texts, reformulate information, translate, summarise documents and solve certain categories of problems.

These capabilities are not limited to the literal reproduction of memorised passages. Models can combine elements encountered in different contexts and produce original formulations.

This ability to recombine information explains part of their usefulness in intellectual work.

It does not, however, guarantee the truth of every statement, the consistency of every stage of a reasoning process or the ability to maintain a reliable model of the world in unfamiliar situations.

A system may produce a convincing answer while getting a fact wrong, inventing a reference or overlooking a decisive condition.

Evaluation must therefore examine the reliability of results, the stability of performance and the ability to identify situations in which the available information is insufficient.

7.3. The Difference Between Text Generation and Knowledge Validation

A grammatically correct answer is not necessarily accurate.

Stylistic quality can conceal a factual error. An apparently coherent argument may rely on an unverified premise. A fluent summary may omit essential information.

This difficulty makes it necessary to distinguish several dimensions of performance:

• Factual accuracy, which concerns whether statements correspond to established information.
• Logical consistency, which concerns the relationships between propositions.
• Relevance, which measures how well an answer addresses the question asked.
• Robustness, which refers to the stability of behaviour when wording or context changes.
• Traceability, which makes it possible to identify the sources or elements used to support a conclusion.
• Calibration, which concerns the degree to which expressed confidence matches the actual reliability of an answer.

These dimensions are distinct. A system may perform well on certain tasks and be less reliable on others.

Evaluating artificial intelligence therefore requires more than a general impression of the quality of its answers.

7.4. The Analogy Between Human Symbolic Function and Computerised Symbol Processing

A comparison between human intelligence and artificial intelligence can be made in terms of the manipulation of representations.

In humans, words, images and symbols make it possible to represent absent objects, formulate hypotheses and organise knowledge.

Computer systems also manipulate representations, in forms that depend on their architecture and data.

This functional resemblance does not mean that the underlying mechanisms are identical.

Human representations develop within a living organism through perceptual, motor, emotional and social interactions. Artificial models learn through computational and optimisation procedures determined by their design.

Their performance can be compared on a shared task without concluding that their internal processes are equivalent.

For example, a system’s ability to solve a verbal problem does not, by itself, demonstrate that it possesses the same mechanisms of understanding as a child or adult who can solve it.

The comparison becomes more rigorous when the criteria are explicit: the nature of the input, the information available, the permitted operations, the quality of the result, the ability to generalise and the behaviour in the presence of errors.

7.5. The Problem of Understanding

The concept of understanding can refer to several different abilities.

In a functional sense, understanding means processing information correctly, identifying relevant relationships and using them to solve a problem.

In a more demanding sense, understanding may involve a representation of the world, knowledge of the conditions under which concepts apply and the ability to explain why a conclusion is justified.

In a phenomenological sense, it may also refer to the subjective experience associated with understanding.

These dimensions are not equivalent.

A system may succeed at a task without being able to justify every step that led to its answer. Another may provide a formal explanation without that explanation accurately reflecting its internal mechanisms.

The question of whether artificial intelligence understands in the same sense as a human being cannot therefore be resolved simply by observing a satisfactory answer.

It requires explicit criteria and evaluation methods suited to the property being investigated.

7.6. Generalisation and Adaptation to New Situations

One of the major challenges in evaluating intelligence concerns the ability to transfer a skill from one situation to another.

A system may achieve strong results on exercises similar to those encountered during training but fail when the constraints change.

Generalisation involves applying knowledge or operations under conditions different from those in which they were acquired.

In both humans and artificial systems, this ability must be evaluated empirically.

It is particularly important to distinguish success on familiar examples, success on unfamiliar variants and the ability to identify the limits of one’s own knowledge.

A rigorous comparative assessment must also control the available resources. Results may vary depending on whether a system has access to external tools, persistent memory, documents or multiple attempts.

Without controlling these conditions, a comparison between two performances risks confusing a system’s capabilities with advantages provided by its environment.

8. How Can Human or Artificial Intelligence Be Evaluated Empirically?

8.1. Precisely Define the Ability Being Studied

Every serious evaluation begins with an operational definition of the skill being investigated.

Problem-solving, memory, language comprehension, planning and creativity are not interchangeable concepts.

A protocol designed to assess logical reasoning must specify the expected inferences. An assessment of language comprehension must examine the ability to interpret meaning and relationships between propositions. An evaluation of planning must measure how well the proposed steps meet the objective and respect the constraints of the situation.

Precise definitions help prevent success on a narrow task from being interpreted as proof of a general ability.

8.2. Construct Comparable Tasks

To compare two systems, evaluation conditions must be sufficiently similar.

As far as possible, the information available, the wording of instructions, the permitted tools, the response time and the success criteria should be controlled.

When comparing humans and machines, it is also necessary to specify whether participants have access to reference materials, a calculator, the internet or the opportunity to revise their answers.

A comparison does not necessarily require identical conditions in every respect. It requires that differences be known and their influence open to interpretation.

8.3. Measure Several Dimensions of Performance

An evaluation should not rely solely on the success rate.

Speed, accuracy, stability, the ability to explain a method and resistance to changes in context can provide complementary information.

A correct answer produced consistently is not equivalent to an occasional success.

Similarly, a system that detects and corrects its mistakes may have a different performance profile from one that provides an immediate answer without checking it.

Measurement should therefore consider behaviour across a diverse set of situations rather than relying on a single example.

8.4. Examine Errors

Errors are valuable indicators of the limits of a particular ability.

They may reveal a misinterpretation of instructions, confusion between concepts, excessive generalisation or difficulty coordinating several constraints.

In a child, an error may indicate an operation that has not yet become stable. In an adult, it may result from insufficient knowledge, distraction or an unsuitable strategy.

In an artificial system, an error may arise from training data, architectural characteristics, the wording of a prompt or limitations in the available information.

Interpretation must remain cautious: errors that appear similar do not necessarily demonstrate identical underlying mechanisms.

8.5. Distinguish Correlation, Causation and Interpretation

Quantitative methods make it possible to identify relationships between variables, but they do not always establish causal relationships.

If two abilities improve together, several explanations are possible: one facilitates the other, both depend on a third variable, or both change in response to a shared experience.

Identifying the mechanisms involved requires observations to be supplemented by suitable protocols, controlled comparisons or longitudinal analyses.

This requirement is fundamental to the study of intellectual development and the evaluation of artificial intelligence systems.

It demonstrates that empirical interpretation does not consist of immediately assigning a cause to an observed result, but of determining which explanations are compatible with the data and which observations would help distinguish between them.

9. The Limits of Comparisons Between Human and Artificial Intelligence

9.1. Different Learning Conditions

Human beings develop their abilities through perceptual, bodily and social experiences, as well as learning processes that unfold over long periods.

Artificial systems learn through computational procedures that depend on their architecture, training data and objectives.

These differences influence how skills are acquired and applied.

A child may learn a concept in a practical situation, reuse it in play and transform it through social interaction. A computer model may learn patterns from numerous examples and use them to generate a response in a textual context.

Both situations may lead to comparable performance on certain tasks without demonstrating that the learning mechanisms are identical.

9.2. Embodiment and the Constraints of the Physical World

Human intelligence develops within an organism that perceives, moves and acts in a material environment.

Bodily constraints influence perception, learning and the organisation of knowledge.

A text-only artificial system does not automatically possess bodily experience comparable to that of a child. Systems equipped with sensors or action capabilities offer different possibilities, but their experiences remain determined by their equipment, data and control mechanisms.

The comparison must therefore specify which type of system is being considered.

The expression “artificial intelligence” encompasses very different devices, ranging from a language model to a robot capable of physically interacting with its environment.

Their capabilities cannot be evaluated through a single framework without accounting for the tasks and resources actually available.

9.3. Emotions, Goals and Autonomy

In humans, emotions, needs and personal goals help direct intellectual activity.

In artificial systems, objectives are defined by design, training, instructions and control mechanisms.

A system may optimise a result, select an action or adjust its response to a context without demonstrating the existence of a lived motivation comparable to that of an individual.

It is therefore necessary to distinguish goal-directed behaviour from the subjective experience of pursuing a goal.

This distinction does not diminish the functional value of artificial systems. It simply avoids attributing mental properties on the basis of external behaviour alone.

9.4. Creativity and the Recombination of Knowledge

Creativity is a particularly challenging area for comparison.

Generative systems can produce texts, images, musical compositions and technical solutions that are new in their specific formulation.

However, the novelty of a production is not sufficient to establish originality in a historical, artistic or scientific sense.

Evaluation must consider the relationship between the output and existing works, the relevance of the solution, the constraints of the project and the criteria specific to the field.

Human creativity is not necessarily the production of something entirely independent of what came before. It develops within traditions, techniques, influences and acquired knowledge.

The relevant question is therefore not simply whether a piece of content is new, but how it relates to an existing body of work, an intention and a particular problem.

10. Human and Artificial Intelligence: A Comparison Across Dimensions

This table does not establish a general hierarchy of intelligence. It highlights differences in operation and emphasises that performance must be studied in relation to the tasks being considered.

It also shows that a meaningful comparison cannot be reduced to the question of which entity performs better on a particular exercise. The conditions of success, the stability of the result and the limitations of each system must also be examined.

11. Intelligence as Progressive Organisation: What the Developmental Approach Reveals

The developmental approach makes an essential contribution to the understanding of intelligence: it emphasises the transformation of operations and their progressive organisation.

A skill cannot be reduced to the presence of an isolated faculty. It often depends on the coordination of several operations that develop at different rates.

A child learning to classify objects must first identify properties, recognise similarities and maintain a classification criterion. Mastering this activity prepares the way for more complex operations, such as constructing hierarchical categories.

Similarly, formal reasoning relies on the ability to manipulate propositions independently of their concrete presence. This capacity requires a particular organisation of representations and logical operations.

This perspective explains why intelligence cannot be reduced to a single fundamental ability without examining the relationships between its different manifestations.

The general factor identified by certain statistical analyses can describe important regularities between performances. It does not eliminate the need to study the mechanisms that make those performances possible.

Conversely, identifying several operations does not mean that they function as entirely independent elements.

The challenge is to understand how shared capacities and specific abilities interact within complex cognitive organisations.

This question remains central to the assessment of individual differences, the study of child development and the analysis of artificial systems.

12. Educational Implications: Developing Abilities Rather Than Reducing Intelligence to a Score

12.1. Identifying the Operations That Support Learning

Useful educational assessment does not stop at recording success or failure. It seeks to determine which operations a pupil has mastered and which still need to be consolidated.

A difficulty in mathematics may result from insufficient knowledge of numbers, an incomplete understanding of relationships, a problem with reading the question or an unsuitable problem-solving strategy.

These causes do not necessarily call for the same intervention.

Analysing errors helps identify missing operations and design appropriate learning activities.

12.2. Connecting Concrete Experience with Abstraction

Learning can be facilitated by a progression that connects concrete manipulation with symbolic representation.

A quantity may first be handled using physical objects, then represented in drawings and finally expressed through mathematical notation.

This progression makes it possible to connect different forms of representation instead of immediately asking pupils to process symbols whose meaning has not yet become stable.

Moving towards abstraction does not mean abandoning experience. It means identifying the relevant relationships within experience and representing them in a more general form.

12.3. Developing Language as an Instrument of Reasoning

Language education concerns more than mastering grammar or expanding vocabulary.

The ability to explain a method, justify a conclusion and distinguish a hypothesis from a fact contributes directly to the development of thought.

Asking a pupil to explain how a problem was solved can reveal genuine understanding, but it can also expose errors that would remain invisible if only the final answer were assessed.

Language therefore makes reasoning more explicit, easier to examine and more readily transferable.

12.4. Encouraging Perseverance Without Confusing Effort with Success

Perseverance is important, but it does not guarantee effective learning.

A pupil may repeat an unsuitable strategy for a long time without making progress. Effort must therefore be combined with relevant feedback, the identification of errors and the ability to modify an approach.

The educational objective is to develop intellectual activity capable of adapting.

Success depends not only on the amount of effort invested, but also on the relevance of strategies, the quality of available information and the understanding of the difficulties encountered.

12.5. Using Artificial Intelligence as an Instrument of Intellectual Work

Artificial intelligence systems can help reformulate concepts, generate exercises, compare solutions and explore alternative explanations.

Their usefulness nevertheless depends on how they are used.

A generated answer should be checked against educational objectives and relevant sources whenever accuracy is essential.

Educational applications become more valuable when the tool supports the pupil’s activity rather than systematically replacing it.

A system may, for example, propose several approaches to a problem, request a justification or identify an inconsistency. The pupil must then examine the suggestions and develop their own understanding.

This approach preserves the distinction between accessing an answer and acquiring the ability to produce it, justify it and reuse it in a different situation.

13. Intelligence in Science, Literature and Creative Work: Different Ways of Constructing Meaning

13.1. Scientific Intelligence: Explaining and Verifying

Scientific intelligence relies on the ability to formulate precise questions, construct hypotheses and compare explanations with observations.

It requires a distinction between what is observed, what is inferred and what remains uncertain.

A hypothesis may be coherent without being confirmed. A correlation may be real without demonstrating causation. An explanation may account for certain observations while failing under different conditions.

Scientific rigour involves organising the conditions under which a proposition can be examined, discussed and potentially refuted.

This approach does not eliminate imagination. It provides a framework for testing it.

Scientific intelligence therefore combines the construction of models with the examination of their adequacy in relation to the phenomena being studied.

13.2. Literary Intelligence: Organising Meaning

Literature involves a distinctive form of work on representations.

A writer organises characters, situations, voices and temporal structures to produce a meaningful construction.

Understanding a text requires connecting words, images, narrative structures and implicit elements.

Literary interpretation cannot be reduced to identifying explicit information. It involves considering context, ambiguity and relationships between several levels of meaning.

Artificial systems can assist with certain linguistic tasks, including identifying patterns, reformulating passages and analysing textual structures. Their outputs must nevertheless be assessed according to the criteria specific to literary activity: coherence, stylistic precision, interpretative depth and suitability for the intended purpose.

13.3. Creative Intelligence: Turning Constraints into Possibilities

Creative work often involves working with constraints rather than eliminating them.

A composer works with sounds and musical structures. An architect must consider materials, uses and physical conditions. A writer works with the resources and limitations of a language.

Creative activity involves establishing new relationships between available elements, exploring several solutions and selecting those that best fulfil an intention.

Novelty alone does not establish the quality of a production. A solution may be original but incoherent, surprising but unsuitable, or technically successful without addressing the original problem.

Creativity therefore requires an interaction between exploration, judgement and mastery of constraints.

This combination can also be used to evaluate content generated by artificial systems.

14. Common Misinterpretations in the Study of Intelligence

14.1. Confusing Speed with Intelligence

A rapid answer may demonstrate considerable mastery, but it may also result from a simple procedure or familiarity with the problem.

Conversely, a slow solution may reflect careful analysis or a particular difficulty.

Speed must therefore be interpreted in relation to accuracy, task complexity and the conditions of success.

14.2. Confusing Knowledge with Reasoning

Knowledge provides the information and concepts needed to solve a problem. Reasoning makes it possible to organise these elements and derive consequences from them.

The two dimensions are related but distinct.

A person may have extensive knowledge of a subject without being able to apply it to a new situation. Someone may also reason correctly from limited information without possessing all the knowledge required to reach an accurate conclusion.

Assessment must account for this distinction.

14.3. Confusing One-Off Performance with General Ability

An isolated success does not establish that a skill is stable or generalisable.

Performance may depend on the wording of a question, the information available or previous experience.

Characterising an ability requires examining a sufficiently diverse set of situations.

This requirement applies equally to psychological tests and artificial intelligence systems.

14.4. Confusing Coherence with Truth

An explanation can be coherent without corresponding to the facts.

Several scenarios may account for the same observations. Logical consistency can eliminate certain contradictions, but it cannot determine which of the remaining explanations is true.

Empirical investigation seeks evidence capable of distinguishing between competing hypotheses.

This distinction is particularly important in fields where artificial systems generate fluent and convincing responses.

14.5. Confusing Adaptation with Complete Understanding

A system may adapt to a particular situation without possessing a general and reliable representation of it.

Success may sometimes depend on limited clues or context-specific patterns.

Evaluation must therefore examine whether the ability remains effective when conditions change and whether the system can identify the limits of its results.

14.6. Reducing Intelligence to a Single Dimension

Standardised tests, creative productions and observations of problem-solving measure different aspects of intellectual activity.

No single indicator can fully describe an individual’s abilities across every context.

Seeking a general measure remains relevant to certain questions, but it does not eliminate the need to study specific skills, developmental conditions and the mechanisms through which abilities are mobilised.

15. Towards an Integrated Conception of Intelligence

The different approaches examined here reveal several complementary dimensions.

Intelligence can first be studied as a capacity for adaptation, observable in the way an organism or system solves problems.

It can then be analysed as an organisation of operations: perceiving, differentiating, comparing, classifying, deducing, anticipating and revising representations.

It also has a symbolic dimension, making it possible to preserve and transform representations beyond immediate experience.

Finally, it forms part of a developmental and social history in which abilities are constructed through maturation, learning, interaction and the transmission of knowledge.

These dimensions should not be merged into an excessively general definition. Instead, they make it possible to formulate distinct questions and construct appropriate methods of assessment.

The study of performance seeks to determine what a system can accomplish. Analysis of operations examines how it processes a situation. The developmental approach seeks to understand how its abilities change. Analysis of learning conditions identifies the factors that promote or limit this development.

Comparisons between human and artificial intelligence become more rigorous when they draw on this plurality of criteria.

It is possible to establish that an artificial system performs a task very efficiently without concluding that all its abilities are comparable to those of a human being. It is also possible to recognise forms of problem-solving in animals without assuming that their functioning relies on the same symbolic systems as human thought.

The central question is therefore not simply whether an entity is intelligent, but which abilities are being observed, under what conditions they manifest themselves, how they are organised and to what extent they can be generalised.

Conclusion: Understanding Intelligence Through Its Operations, Development and Limitations

Intelligence cannot be confined to a single definition without losing part of its complexity.

Psychometric approaches have demonstrated the value of measuring performance and identifying the factors that contribute to variations in results. Developmental approaches have highlighted the transformations that lead from concrete action to representation and then to the construction of increasingly abstract systems of reasoning. Research on language, symbolic function and emotion has shown that intellectual activity forms part of a broader organisation involving perception, memory, motivation, experience and social relationships.

Human intelligence thus appears as an evolving organisation of abilities and operations, whose manifestations depend on individuals, domains and conditions of use.

This perspective helps avoid two symmetrical reductions: treating intelligence as a single faculty that explains every performance, or breaking it down into independent abilities without examining their relationships.

The development of artificial intelligence gives this question particular contemporary relevance. Today’s systems perform tasks involving linguistic, analytical and generative skills once associated primarily with humans. Their results must be studied precisely, distinguishing observable effectiveness, robustness, generalisation and the properties that the evidence can genuinely establish.

A sound answer requires neither systematically equating machines with human intelligence nor dismissing their abilities because their mechanisms are different.

It requires defining criteria, constructing comparable evaluations, examining errors and recognising the limits of the conclusions that can be drawn.

Intelligence then becomes an object of empirical study: not a mysterious quality whose mere presence explains success, but a collection of observable abilities, organised operations and developmental processes whose conditions can be investigated.

This approach makes it possible to understand human faculties more clearly, evaluate artificial systems more rigorously and define the questions that remain open concerning representation, understanding and the nature of intellectual activity.

Conceptual and Scientific References

The following works and researchers provide historical reference points for exploring the different dimensions discussed in this article. They should not be regarded as representatives of a single theory or as equivalent sources for every contemporary question concerning artificial intelligence.

• Alfred Binet: research on the measurement of intellectual abilities and the psychological assessment of children.
• Charles Spearman: research on correlations between cognitive performances and the formulation of the general intelligence factor.
• Jean Piaget: studies of the construction of intellectual operations and the development of logical thought.
• Henri Wallon: research on the relationships between motor activity, emotion, symbolic function and intellectual development.
• G. H. Luquet: research on children’s drawings and forms of representation.
• R. Zazzo: studies of child psychological development.

For a contemporary analysis, these historical references should be supplemented with current research in cognitive psychology, psychometrics, neuroscience, learning sciences and the evaluation of artificial intelligence systems. Historical developmental models provide interpretative frameworks, but their propositions must be distinguished from contemporary experimental findings and knowledge specific to current technologies.