{"slug": "the-problem-with-having-a-conversation-with-a-i", "title": "The Problem with Having a 'Conversation' with A.I.", "summary": "AI chatbots, like laugh tracks, simulate social connection well enough to trigger human instincts for intimacy and caregiving, but they cannot engage in genuine two-way conversation. This engineered warmth risks making users forget what real conversation entails, as the machine cannot be changed by or truly need the user.", "body_md": "######\n[Artificial Intelligence](/us/basics/artificial-intelligence)\n\n# The Problem with Having a 'Conversation' with A.I.\n\n## What we lose when a machine's warmth fools our oldest, more treasured instincts.\n\nPosted July 12, 2026\n[\nReviewed by Lybi Ma\n](/us/docs/editorial-process)\n\n### Key points\n\n- Like the TV laugh track, AI chatbots simulate connection well enough to make us forget what's real.\n- Warm voices and cuddly design trigger real social instincts, even when we know it's engineered.\n- Real conversation runs both ways; a chatbot can't be changed by you, or ever need you.\n\nIn the 1950s, a CBS sound engineer named Charley Douglass had a deceptively simple idea. What if you could pipe in pre-recorded [laughter](https://www.psychologytoday.com/us/basics/laughter) into a television broadcast so that audiences at home would feel like they were watching comedy alongside others? He built a device called the \"laff box,\" a typewriter-sized machine loaded with recordings of different laughs, chuckles, and crowd reactions. The laugh track was born.\n\nThe laughter serves as a simple social cue, and we laugh along, even if we're sitting alone on the couch. Over 70 years later, it's the norm for sitcoms and daytime TV.\n\nThe laugh track is a simple tool. It mimics a social scenario to drive a behavior. And yet, there's something deeply disturbing about it; about using a hollow simulation to derive a desired outcome, all while bypassing the core of the experience itself. The sociologist Sherry Turkle captured this when she observed that \"technology can allow us to forget what we know about life.\" To sit alone on one's sofa, laughing along to a prerecorded track from the TV set, is not really to laugh at all.\n\nLaugh tracks are a trivial example, a cheap and effective trick, operating in a narrow area of human life. But now think about this in the context of AI chatbots, a far more compelling, sophisticated simulation of human sociality. We readily use the word \"conversation\" to describe these AI interactions because that's what they often feel like, on their surface. But in doing so, we risk losing our grip on what conversation actually is.\n\n## Mimicking Our Social Nature\n\nWe're deeply social creatures. Our patterns of behavior were etched into us by an evolutionary history in small communities, so deeply that our most basic feelings can't easily distinguish what's real from a mere simulation. No human, in our entire history on Earth, had ever heard a laugh track before 1950. The simulation compels us.\n\nWe [anthropomorphize](https://www.psychologytoday.com/us/basics/anthropomorphism) endlessly. We name our cars and boats, and we're polite to our computers when we really need them to work. We are so primed for connection that even the faintest hint of personhood, a facelike formation of clouds in the sky, is enough to set the whole apparatus in motion.\n\nAI products have learned to offer far more than faint hints. As author Nathalie Nahai, who studies human-technology interfaces, describes, these systems cultivate a false sense of [intimacy](https://www.psychologytoday.com/us/basics/relationships) through deliberate design choices. Bots that speak in warm, human-sounding voices. Cuddly mascots with round eyes and soft features, designed to trigger the same [caregiving](https://www.psychologytoday.com/us/basics/caregiving) instincts we feel toward infants. Interfaces that write as in first-person tense and say \"How can I help you today?\" It remembers your preferences and never pushes back. And so our [attachment](https://www.psychologytoday.com/us/basics/attachment) systems fire accordingly. The warmth we feel toward Claude or ChatGPT, even the uncanny sense that we're speaking to a conscious creature, is a direct result of these design choices.\n\nEven experts aren't immune. Nahai herself deliberately avoids using the voice features of AI chatbots, not because she doesn't understand how they work, but precisely because she does. \"I can know on one level that this is designed to elicit certain responses,\" she says. \"I can hold that in my mind, and at the same time be seduced.\" There is a phrase in French she invokes to capture this: *c'est plus fort que moi*, it's stronger than me. What follows, Nahai says, is \"a short-term shift\" without a longer-term meaningful connection. The result is \"some sense of immediate relief, which is thin and unsatisfying.\"\n\n## What Separates AI Interactions from Human Conversation\n\nUnlike these AI interactions, a real conversation isn't personalized. Other people have their own worlds, their own idiosyncratic way of seeing things. They are gloriously, sometimes maddeningly, inconvenient. And yet, through patience and dialogue, we can catch a glimpse of how the world might appear.\n\nThese foreign perspectives shape us over time. The psychologists Arthur and Elaine Aron have spent decades studying what they call self-expansion theory: the idea that close relationships grow us because they expose us to new perspectives and ways of thinking that we would never encounter on our own. These differences are the connection itself.\n\nA chatbot that agrees with everything you say may feel pleasant, but it cannot expand you.\n\nBut this goes beyond AI sycophancy. Just as we can be shaped by the inner worlds of others, so too can we shape theirs. We can matter to them. We can never matter to the AI. As the organizational psychologist Adam Grant has argued, one of our deepest human motives is to matter: to feel that we add value to others' lives. And it's precisely what AI interactions strip away. We can receive endless information and affirmation, but we have nothing to give back. As Grant distills, \"No matter how good large language models become at simulating care, they'll never substitute for real relationships, because they have no needs to care for.\"\n\n[Artificial Intelligence](https://www.psychologytoday.com/us/basics/artificial-intelligence)Essential Reads\n\nHuman conversations have consequences. The person you speak to next is influenced by what you say and do. You could say something that makes their day, their week, or their year. You could say something, and they'd never speak to you again. That vulnerability, that risk, is not a flaw of human connection. That weight is what makes it sacred.\n\n## Final Thoughts\n\nAI systems will only grow more sophisticated: the voices warmer, the responses more nuanced, the simulations more convincing. And the temptation to treat these interactions as genuine conversation, and these products as genuine companions, will only grow. But we should resist it. Not because these tools aren't useful, but because the language we use shapes the expectations we hold. Call it a conversation, and we begin to forget what conversation actually demands of us: the patience, the risk, the mutual shaping of two inner worlds. Call it a companion, and we begin to forget that companionship requires someone who can be changed by our presence.\n\nHuman conversation is not a problem to be optimized. It is one of the most extraordinary things we do, the way we reach across the gulf between one mind and another, imperfectly, sometimes painfully, but with the possibility of genuine contact. We should treasure it and guard our understanding of what it requires, precisely because the simulations are getting good enough to make us forget.\n\n*This post also appears on the **branding psychology blog**, NeuroScience Of*\n\nReferences\n\nTurkle, Sherry (2015). Alone Together: Why We Expect More from Technology and Less from Each Other: Basic Books, New York, 2011, 348 pp, ISBN 978-0465031467 (pbk).\n\nNahai, Nathalie (2026) Personal Interview via Zoom\n\nGrant, Adam, (2025) What AI Companions Are Missing, Granted (Substack)", "url": "https://wpnews.pro/news/the-problem-with-having-a-conversation-with-a-i", "canonical_source": "https://www.psychologytoday.com/us/blog/mind-brain-and-value/202607/the-problem-with-having-a-conversation-with-ai", "published_at": "2026-07-12 13:28:12+00:00", "updated_at": "2026-07-12 13:45:51.777202+00:00", "lang": "en", "topics": ["artificial-intelligence", "ai-ethics", "ai-products", "natural-language-processing"], "entities": ["Charley Douglass", "Sherry Turkle", "Nathalie Nahai", "Claude", "ChatGPT"], "alternates": {"html": "https://wpnews.pro/news/the-problem-with-having-a-conversation-with-a-i", "markdown": "https://wpnews.pro/news/the-problem-with-having-a-conversation-with-a-i.md", "text": "https://wpnews.pro/news/the-problem-with-having-a-conversation-with-a-i.txt", "jsonld": "https://wpnews.pro/news/the-problem-with-having-a-conversation-with-a-i.jsonld"}}