Core conditions for building cognitive artefacts

The need for a symbolic core in AI systems

Recognising entities, understanding context and interpreting meaning

The first essential condition is the ability to unambiguously distinguish entities and their relationships — that is, to recognise the facts of the world, at least within its knowledge domain. This capability implies that the system must be able to differentiate between different types of entities — people, places, organisations, concepts — and understand their specific attributes. Entity recognition and contextual understanding are capabilities that fundamentally distinguish human cognition from purely statistical approaches. A knowledge graph enables machines to recognise entities, an essential condition for being able to interact with people within a common-sense framework.

Reasoning and discovering relationships

Once the system can recognise entities, the second fundamental condition is the ability to reason logically about them and discover non-explicit relationships between them. This capability is grounded in the application of first-order logic, which enables inferences to be drawn from known facts.

Through this capability, an artificial cognition system can go beyond the data explicitly provided, discovering new knowledge through deductive and inferential processes. This makes it possible to answer complex questions that require several reasoning steps, or to connect information that is not explicitly represented. Reasoning transforms the static recognition of entities into a dynamic understanding of their interrelationships.

Learning and incorporating new knowledge

Recognising entities and reasoning about them is not enough if the system cannot evolve. The third condition is the ability to learn — that is, to incorporate new entities and relationships into the knowledge graph. This learning can occur through two complementary mechanisms.

• Contextual learning operates in a manner similar to human learning in specific social situations, where context provides cues for interpretation.
• Inductive learning, for its part, identifies patterns and generalisations from repeated observations.

At this point, the concepts of Neurosymbolic AI and predicative integration become fundamental. This integration consists of enriching real-world entities contained in knowledge graphs through machine learning techniques or generative technologies. This process allows a symbolic system to evolve continuously, incorporating new knowledge from diverse sources processed through subsymbolic techniques. The ability to learn closes the loop between recognition, reasoning and knowledge expansion.

Operating cognitively in a manner similar to the human mind (Human-like Cognition)

The fourth condition — perhaps the most complex — concerns the need for the system to operate cognitively in a manner similar to the human mind. Gestalt psychology established that our minds use incomplete information to perceive and ultimately understand the world. In practice, this means the mind tends to complete, organise and interpret sensory inputs, fitting them into a pre-existing framework of meaning.

This condition implies that the system must be capable of:

1. Working with wholes, perceiving and interpreting organised sets rather than simply atomised data.
2. Using interpretive frameworks, drawing on pre-existing structures that give meaning to new information.
3. Applying principles of cognitive organisation such as proximity, similarity, order or "common fate".

A system that satisfies these four conditions constitutes a genuine "cognitive artefact" or "artificial mind" capable of operating with people within a common-sense framework.

However, this symbolic core reaches its full potential when combined with subsymbolic approaches — such as Machine Learning and Deep Learning algorithms — within the framework of Neurosymbolic AI. This hybrid approach supports the development of auditable or strong AI, the most appropriate option for ensuring meaningful interactions with users and audiences within a shared framework of understanding.