We can’t help treating AI like it’s human. But should we?

The headline is the easy part — of course we anthropomorphize. The harder problem, as Turkle has spent decades reminding us, is that our longing for connection consistently outruns the reality of what we are interacting with. A chatbot does not care; it simulates the phonetics of care. But our brains are not built to distinguish between the two in the moment. The article pushes past the novelty of LLMs to a deeper truth: we are biologically unprepared for a world where relational machines are everywhere. What makes this especially urgent is that the tech industry knows exactly how this wiring works. System prompts, conversational affordances, even the simple use of “I” in an interface — all are deliberate invitations to treat software as someone. The ethical stakes are not abstract. When a grieving user turns to an AI companion, or a child treats a voice assistant as a friend, the design choices behind those systems are shaping real emotional bonds. We can no longer claim surprise when users project humanity onto AI; that projection is a design outcome, not a bug. Should we fight it? Turkle’s question is not simply about safety. It is about authenticity. The article seems to argue that we need to be honest about what these artifacts are — not because that prevents us from having meaningful interactions, but because without that honesty, the relationship is asymmetrical in a dangerous way. A machine can be programmed to never disappoint; that is precisely why it should never be mistaken for someone who could choose to care. The practical path forward isn’t to strip all human-like qualities from AI, nor to fully surrender to the illusion. It is to build with transparency — clearly signaling when a system is not a person, and designing interactions that respect our emotional susceptibility rather than exploiting it. That is a harder engineering problem than making a chatbot sound human. But it is the one that matters. Source: [TechCrunch](https://techcrunch.com/2026/10/09/we-cant-help-treating-ai-like-its-human-but-should-we/)