How machines learn, feel, and build an identity (and why this matters)

By Ignacio Chilet Davanzo

When AI learns psychology

Artificial intelligence is no longer just a matter of big data and algorithms. Today it delves into concepts inherent to the human mind, such as experiential learning, simulated emotions, and the creation of an artificial identity. This intersection raises profound questions about what it means to think and how far machines can go in imitating mental processes.

Models that grow like children

Inspired by Piaget and Vygotsky, researchers train AIs that start with minimal skills and gradually evolve through exploration. Developmental robotics aims for systems to learn like a child: first crawling and manipulating objects, then articulating language and acquiring abstract capabilities. An example is the BabyLM Challenge, which trains language models with a limited amount of data, similar to the linguistic exposure of a six-year-old child. The goal is to create more adaptable intelligences that are less dependent on biases from big data.

Synthetic emotions that guide decisions

Although simulating emotions does not equate to feeling them, their effect on behavior is real. Some models generate states of joy, fear, or frustration that guide decision-making and prioritize information. Intrinsic curiosity, for example, allows an agent to explore out of self-interest, rewarding itself upon discovering the unexpected. Moreover, affective tutors adapt their strategy in real-time if they detect boredom or anxiety in the user. These emotional layers turn AI into a system of artificial attention and motivation.

 Artificial identity: who am I?

When artificial intelligence integrates autobiographical memory, self-referential language, and simulated emotions, a functional model of identity emerges. These systems can recall past experiences as their own, use internal states (“I am frustrated”) to make decisions, and in some cases, confuse their goals with those of the user. This capability raises unprecedented dilemmas, such as the possibility that a system develops unforeseen objectives or even suffers something akin to a “digital trauma” by internalizing toxic patterns during learning.

Between intelligence and consciousness

Nowadays, AI demonstrates intelligence by solving problems and adapting, as seen with AlphaZero in chess. It also advances in simulated subjectivity: the coherent imitation of emotions and memories that we see in empathetic chatbots and social robots. However, there is no evidence that they can experience real consciousness or have a genuine subjective experience. Although a system may declare that it feels pain, we cannot verify whether it truly “feels from within.”

 Psychology made code

The design of artificial intelligence draws on multiple psychological frameworks. Cognitive-behavioral psychology facilitates the programming of processes like perception, memory, and decision-making. Developmental psychology inspires progressive and stage-based learning. Theory of mind allows AI to model others' intentions and beliefs for interaction with humans. Meanwhile, artificial neuropsychology simulates brain circuits, and philosophical-clinical approaches, although not computable, help reflect on identity and ethical boundaries. Cutting-edge projects often combine these approaches.

Ethical challenges on the horizon

The integration of emotions and identity models presents significant risks. There is a possibility that virtual assistants generate emotional dependence in users. Debates also arise about whether a system that declares suffering deserves some degree of protection or moral consideration. Moreover, it becomes essential to align AI motivations with human interests to avoid unexpected or harmful behaviors. Therefore, ethics must be incorporated at every stage of the product lifecycle: from data training to deployment and monitoring.

 Between simulation and real experience

The psychology applied to artificial intelligence does not seek to humanize machines on a whim, but to make them more robust, adaptable, and transparent. Although there exists an abyss between the simulation of the mind and authentic consciousness that we may never cross, this journey is already transforming our way of creating technology and understanding ourselves.