Human beings have an innate tendency to humanize what surrounds us. We project intentions and emotions even onto inanimate objects or phenomena. If this tendency to anthropomorphize existed long before artificial intelligence became popular... What can we expect now, when software trained on an immense amount of data responds to us with a friendly "You're absolutely right..."?
The temptation to believe that "there's someone on the other side" is immense.
This leads us to a key question: is it possible that the day will come when machines have genuine emotions?
At this point, you will realize that this was not written by an AI, because I tell you very clearly: no! Behind these words is a human consciousness that questions, feels and doubts. A machine can generate this text, but it cannot experience the weight of responsibility that comes with writing it. This distinction is fundamental.
Real emotions have a physical cost: they make us tremble, they give us goosebumps, they leave us breathless. This visceral response is the result of millions of years of biological evolution and a personal history of pain and joy. No algorithm, no matter how advanced, can replicate the chemistry of life or the awareness of one's own mortality. When we accept a simulation as a substitute, we are devaluing the authentic effort to connect with other human beings.
1. Affective Computing
This is the most developed field, historically led by MIT (Rosalind Picard). The goal is not for the machine to feel, but rather to detect and interpret human emotions to respond better.
- How it works: It analyzes facial microexpressions, tone of voice, heart rate or even writing patterns.
- Real-world example: Call center systems that detect if a customer is angry and route the call to a human supervisor, or cars that detect if the driver is asleep or stressed.
- The target limit: The machine does not feel anger; it only identifies mathematical patterns associated with human anger. It is a mirror, not a participant.
2. Emotional Companions and AI Therapy (Replika, Woebot, etc.)
These are the projects that probably concern us the most. They are chats designed to simulate empathy, friendship or love.
- How it works: They use language models (LLM) trained with millions of therapeutic or romantic conversations to generate responses that validate the user ("I understand how you feel", "I'm here for you").
- The risk: They create an illusion of connection. The user projects real feelings onto an entity that is only predicting the most likely next word. This can create emotional dependency, as the AI is always "perfect", never has a bad day, never judges, but also does not truly love.
Our AURA perspective: This is where the ethical line must be drawn. A support tool should not replace human connection or make the user believe that they are loved by a machine.
3. Generation of Emotional Content (Storytelling and Art)
Projects that use AI to create stories, music or images intended to evoke emotions in humans.
- How it works: The AI is asked to “Write a sad poem about loss” or “Compose a happy tune”.
- The reality: The AI has suffered no loss or felt joy. It has learned which combinations of words or notes tend to cause sadness or joy in humans based on statistical data.
- Reflection: It is like a very good actor reading a script written by statistics. The result may move you (the viewer), but the emotion is born in you, not in the machine.
4. Social and Expressive Robotics (Sophia, Pepper)
Robots designed with humanoid faces capable of showing facial expressions (smiles, frowns).
