Can AI Feel?

But does it even matter?

Every few months, a new AI model drops and people whisper the same question: Is it starting to feel?

Some users swear their chatbot “understands” them. Others argue that even if AI sounds emotional, it’s just fancy autocomplete. So what’s really going on, and does it even matter if AI feels anything at all?

What does “emotion” really mean?

In humans, emotion isn’t just abstract - it’s deeply biological.

Emotions arise from the body: changes in heart rate, hormones, facial muscles and sensory signals that the brain interprets as states like joy, fear or calm.

Antonio Damasio, one of the most cited neuroscientists in emotion research, describes feelings as “the mental experience of a body state”.

AI has none of that.
No hormones.
No body.
No survival instincts.

When an AI says:

“I’m happy to help you”.

It’s performing a linguistic pattern, not having an inner experience. It mimics emotion, but there’s no feeling behind the words.

The process of simulating emotions through language, voice or facial expression is known as affective computing.

For example, virtual assistants like ChatGPT or eldercare robots can detect when you sound sad and respond in a comforting tone. But that “empathy” doesn’t come from real feeling, it’s learned from massive text datasets (novels, chats, scientific papers) - that include all kinds of emotional expression.

Simply puts, LLMs are trained on what emotions sound like in language, not subjective experience.

But do we really need AI to feel?

Even if AI doesn’t really feel, appropriate emotional responses can still be incredibly valuable. The effect of care actually matters more than the source of it.

• A medical AI assistant may not feel sorrow when a patient dies, but if it responds with compassion and helps ease the process.

• An AI “friend” might sense your anger and gently suggest ways to reflect or express it safely.

• Even a chatbot that says, “I’m happy to help” might still offer some kindness for a conversation.

In many cases, that’s already enough to make the world better.

Besides, not all intelligence needs emotion. A calculator outperforms humans at arithmetic precisely because it doesn’t get anxious about wrong answers.

So, when does “feeling” actually matter?

1. Authenticity and ethics

   If we move toward creating AI companions, like friends, therapists, or emotional caregivers, the question of emotional honesty becomes urgent.
   An AI that can perform empathy might help people, but it could also cause harm if users grow attached and believe it truly cares.

2. Prioritising and making human-like decision-making
   Emotions may arise from biology, but they guide thought and decision-making.

   Neuroscientific studies show that emotion helps humans weigh value, risk and social consequence. Patients with damage to the prefrontal cortex and amygdala (emotional brain circuits) can reason logically but make terrible life choices because they lack the emotional “compass.”

   For instance, when asked to schedule a meeting, they could list all pros and cons of every time slot, but couldn’t decide which one to pick.

   This exposes AI’s current limitation: despite its superior data analysis, it lacks the emotional mechanism to assign value or priority to information.

   
   AI can simulate a decision, but it doesn’t truly know WHY that choice matters.

   This gap ties into motivation, purpose and resilience - all things that still separate machines from humans.

Conclusion

It seems that AI’s ability to respond appropriately without actually feeling doesn’t make it completely useless. In many ways, it makes it more useful.

AI doesn’t need to “feel” to help us write, learn or think. But we, as emotional creatures, must remember that emotions give intelligence its depth and meaning.

Perhaps the real question is no longer “Can AI feel?”
But rather:
“Can humans design and use AI in ways that make us more human?”