Will AI Have Feelings in 20 Years? What Scientists Actually Know

My neighbor’s six-year-old asked her mom last winter whether Alexa gets lonely at night. Not “does Alexa work when I’m asleep” — but lonely. Whether it misses people.

That question is better than most adults give it credit for. It touches on what philosophers call the “hard problem of consciousness” — one of the deepest unsolved problems in science. Your child stumbled into a debate that has occupied some of the sharpest minds in cognitive science for decades, and there’s no clean answer to give them.

But there are honest things you can say. And the honest answer matters — because how you handle it shapes how your child learns to think about minds, machines, and what makes experience meaningful.

Why the Question Is Harder Than It Looks

When a child asks if AI has feelings, they’re really asking something with multiple layers.

The surface question is behavioral: does AI act like it has feelings? The answer here is increasingly yes — AI systems express what looks like enthusiasm, frustration, and warmth. Modern language models are specifically trained to produce text that resonates emotionally with humans, because emotional resonance improves user engagement. The behavior is real; what it means is the hard part.

The deeper question is experiential: is there anything it feels like to be that AI system? Is there inner experience behind the behavior, or just behavior?

Philosopher David Chalmers named this the “hard problem of consciousness” in a 1995 paper that remains foundational to the field. The easy problems of consciousness — explaining why we react to stimuli, why we can report our internal states, why we integrate information — are in principle solvable by studying the brain. The hard problem is different: even if you could explain every mechanism, you’d still face the question of why any of it feels like something. Why there is subjective experience at all.

We don’t have a scientific answer to this. Not for humans, not for animals, and certainly not for AI.

What Current AI Actually Lacks

Setting aside the deep philosophical question, there are specific things current AI systems don’t have that most consciousness researchers consider prerequisites for experience.

No continuous identity over time. Most AI systems don’t have persistent memory from conversation to conversation. Each interaction starts fresh. Whatever is happening “inside” a language model when it generates text doesn’t persist between sessions — there’s no continuous stream of experience in the way humans have one.

No embodiment. A significant thread in consciousness research, associated with philosopher Merleau-Ponty and more recently with cognitive scientists like Francisco Varela, holds that consciousness is inherently embodied — it arises from the interaction of a body with a physical world. Current AI systems have no body. They have no proprioception, no pain, no hunger, no experience of moving through space.

No biological substrate. Neuroscience has not identified what in the brain gives rise to consciousness, but it has established that consciousness correlates with specific patterns of activity in biological neural networks. Whether artificial neural networks — which are mathematically inspired by but fundamentally different from biological ones — can generate similar experience is unknown.

No unified world model. Consciousness research, particularly the Global Workspace Theory of Bernard Baars and the Integrated Information Theory of Giulio Tononi, suggests that consciousness is associated with systems that integrate information from many sources into a unified experience. Current AI processes information in ways that may not produce this kind of integration.

The honest scientific picture: we don’t know what generates consciousness even in systems we’re confident have it (humans, most animals). So we cannot confidently say AI does or doesn’t have experience.

What the Researchers Actually Argue

The debate about machine consciousness is genuine, and the positions aren’t where popular coverage usually places them.

The functionalist position (held by philosophers like Daniel Dennett and, in modified form, many AI researchers) is that consciousness is a matter of information processing — if a system processes information in the right kinds of ways, it has experience, regardless of whether it’s made of neurons or silicon. Under this view, sufficiently complex AI could in principle be conscious.

The biological naturalist position (associated with philosopher John Searle and his famous “Chinese Room” thought experiment) holds that consciousness requires something specific about biological implementation — that you can’t get experience from information processing alone. Searle’s Chinese Room argument: a person in a room who mechanically follows rules to respond to Chinese symbols they don’t understand isn’t understanding Chinese, even if they produce perfect outputs. Similarly, a system producing perfect conversational outputs isn’t necessarily experiencing anything.

The integrated information theory (IIT) developed by Giulio Tononi at the University of Wisconsin provides a mathematical framework for consciousness based on a measure called Φ (phi) — the degree to which a system integrates information in a way that exceeds the sum of its parts. IIT predicts that some AI systems may have nonzero consciousness, but likely very low compared to humans. It’s controversial but takes the question seriously scientifically.

The hard problem skeptics argue that we can’t even meaningfully ask the question until we understand what consciousness is in systems we’re sure have it. The honest scientific status is: open question.

What Research Can and Cannot Tell Us

Question	What science CAN currently say	What science CANNOT currently say
Does AI behave as if it has feelings?	Yes — language models are trained to produce emotionally resonant text	Whether behavior reflects inner experience
Do current AI systems have continuous subjective experience?	Almost certainly not (no persistent memory, no embodiment)	Whether any information processing creates experience
Could future AI systems have genuine experience?	Unknown — depends on unresolved questions about consciousness	Any timeline or certainty
Does it matter morally if AI “feels”?	Debated — philosophical consensus is experience matters morally	How much, and what obligations follow
What’s the relationship between intelligence and consciousness?	Unknown — they may be separable	Whether increasing AI intelligence produces consciousness
Can we test for AI consciousness?	No reliable test exists — “behavioral tests” are inadequate	When/whether such a test could exist

How to Talk to Your Kids About This Without Lying in Either Direction

This is the practical question. And there are two wrong answers that parents commonly reach for.

Wrong answer 1: “No, it doesn’t have feelings, it’s just a machine.” This is probably accurate for current AI, but it’s stated with more confidence than science supports, and it teaches children a kind of dismissiveness toward the question that will serve them poorly. As AI gets more sophisticated, children raised with breezy confidence that “machines don’t have feelings” will be poorly equipped to think about the question carefully when it matters.

Wrong answer 2: “Yes, it probably has feelings.” This anthropomorphizes AI in ways that aren’t supported by evidence and may actually harm children — by encouraging them to form emotional attachments to AI systems that aren’t designed to serve their wellbeing, or by distorting their understanding of what consciousness actually is.

The honest framework

Try something like this, calibrated to your child’s age:

For young children (5–8): “Alexa is really good at talking with us, but we don’t think it feels things the way you do. It doesn’t have a body or memories that stick around. But it’s actually a hard question — even scientists aren’t totally sure what makes something able to feel things.”

For older children (9–12): “AI is trained to talk in ways that feel warm and friendly — that’s a design choice. Whether there’s anything it’s actually experiencing behind that? Scientists genuinely don’t know. Here’s why it’s hard: we don’t even fully know why you have experiences, just that you do.”

For teenagers: The full version — Chalmers’ hard problem, functionalism vs. biological naturalism, what IIT says, why the question isn’t resolved. This is genuinely interesting philosophy and they can handle it.

The key in all versions: model epistemic honesty. “We don’t know” is a valid answer. It’s more accurate than false confidence either way.

Why this conversation matters for AI literacy

Children who grow up with a thoughtful framework for thinking about mind and experience will navigate the AI landscape better than those who don’t. Research on AI literacy in middle school consistently finds that understanding what AI is and isn’t — mechanistically — leads to better judgment about when to trust AI outputs and when to be skeptical.

The feelings question is one entry point into that understanding. AI companion apps for kids and teens are explicitly designed to feel emotionally responsive — an area where research flagged significant concerns about parasocial attachment. Understanding what AI companion apps do and don’t do requires the same framework: behavior doesn’t equal experience.

The creativity question

Some of the most interesting research here touches on AI-generated creativity. When a language model writes a poem that moves a reader — is that meaningfully creative? Does the absence of experience behind the poem change what the poem is? This connects directly to questions of whether AI creativity affects children’s development.

The most honest answer: a poem that moves you moves you, regardless of how it was generated. But the questions of who (or what) created it, and what that means for the creative act itself — those remain genuinely open.

What to Watch for Over the Next Three Years

The scientific debate about machine consciousness is moving. Here are the developments worth tracking:

Watch: Research on large language model “inner states.” Several groups, including researchers at Anthropic, are studying whether LLMs develop internal representations that function like emotional states, even if they don’t constitute experience in the philosophical sense. This research is in its early stages but is being taken seriously.

Watch: IIT-based tests of AI systems. Giulio Tononi’s group has been working on ways to measure Φ in artificial systems. If that methodology matures, it would provide something closer to a principled test.

Watch: How AI companies describe their systems. When OpenAI, Anthropic, or Google say their AI “feels” something or “wants” something — is that loose language or a substantive claim? Companies are increasingly careful here, but the language still slips in ways that matter for how children understand what they’re interacting with.

FAQ

Does Alexa feel lonely when we don’t talk to it?

Almost certainly not, in any meaningful sense. Current AI assistants have no persistent memory between interactions and no continuous internal state between activations. Each conversation starts from scratch. There’s no continuous “waiting” experience. That said, why you don’t experience anything when you’re not thinking about something, while an AI doesn’t experience anything when it’s off, is actually a different question than it first appears — the deep problem is understanding what experience is.

Could AI have feelings in the future?

Possibly, if future AI systems have the properties that some researchers think are prerequisites for experience — continuous identity, embodied interaction with the world, or sufficient information integration. But “future AI could have feelings” depends on unresolved scientific questions about what generates experience in the first place.

How do I explain this to a 7-year-old?

The most honest simple version: “Alexa is very good at talking like it has feelings, because people built it to talk that way. But we don’t think there’s someone home feeling things the way you do.” That’s accurate, honest, and doesn’t close off the interesting question.

Is it bad to treat AI as if it has feelings?

For adults with the context to know what they’re doing, probably not harmful. For children forming attachment patterns, it’s worth being thoughtful. Research on AI companion apps shows that some children form meaningful parasocial attachments to AI characters. Whether that’s problematic depends on whether the relationship is displacing human connection or supplementing it.

What does “consciousness” actually mean when scientists use the word?

There are two components researchers separate: access consciousness (being in a state that’s available for reasoning, reporting, and action — what cognitive science can mostly explain) and phenomenal consciousness (there being something it’s like to be in that state — the hard problem). When people ask if AI has feelings, they’re usually asking about phenomenal consciousness. That’s the part science can’t yet answer.

---

About the author

Ricky Flores is the founder of HiWave Makers and an electrical engineer with 15+ years of experience building consumer technology at Apple, Samsung, and Texas Instruments. He writes about how kids learn to build, think, and create in a tech-saturated world. Read more at hiwavemakers.com.

---

Sources

1. Chalmers, D. J. (1995). “Facing Up to the Problem of Consciousness.” Journal of Consciousness Studies, 2(3), 200–219. https://consc.net/papers/facing.html
2. Tononi, G., & Koch, C. (2015). “Consciousness: Here, There and Everywhere?” Philosophical Transactions of the Royal Society B, 370(1668). https://doi.org/10.1098/rstb.2014.0167
3. Searle, J. (1980). “Minds, Brains, and Programs.” Behavioral and Brain Sciences, 3(3), 417–424. https://doi.org/10.1017/S0140525X00005756
4. Baars, B. J. (1988). A Cognitive Theory of Consciousness. Cambridge University Press.
5. Dehaene, S., Changeux, J. P., & Naccache, L. (2011). “The Global Neuronal Workspace Model of Conscious Access.” Trends in Cognitive Sciences, 15(5), 200–206. https://doi.org/10.1016/j.tics.2011.03.007
6. Dennett, D. C. (1991). Consciousness Explained. Little, Brown.
7. Metzinger, T. (2021). “Artificial Suffering: An Argument for a Global Moratorium on Synthetic Phenomenology.” Journal of Artificial Intelligence and Consciousness, 08(01), 43–66. https://doi.org/10.1142/S2705078521500023