People are polite to machines, and the AI industry is building products around that habit. A new TechCrunch AI feature looks at why we treat AI like it’s human. It starts with a reporter at MIT’s CSAIL greeting a $50,000 Unitree humanoid named Gabe with “Hello, sir, it’s nice to meet you” and a handshake. She knew the robot wasn’t sentient. She still wanted to be its friend.
That impulse is the story. As detailed in TechCrunch AI, the researchers who study it most closely say it’s built into us. They also say tech companies are only starting to answer for how they use it.
🧠 We’re wired to care back
Dr. Sherry Turkle has studied how people relate to computers since joining MIT in 1976. In her new book, Artificial Intimacy, she sums up the problem: “When we are drawn into even the most primitive exchanges with a relational artifact, we believe it cares for us. And we are wired to care for it in return.”
The data backs her up:
- One study found about 70% of people are polite when they interact with AI.
- Another found users get more likely to say “please” or “thank you” once a chat goes past the first prompt and response.
So the longer the conversation runs, the more human the machine feels. That’s a big deal for an industry pushing AI into search, customer service and companion apps, where long, sticky sessions are the goal.
⚖️ Two views on where the line is
The MIT researchers don’t fully agree, and the difference is useful.
Turkle draws lanes. She isn’t against AI. She once advised a student who built a bot to help people practice job interviews. “Stay in your lane, chatbot!” she said at her book launch. “Coaching for a job interview? That’s a lane.” A chatbot that tells a three-year-old to talk to it instead of their parents “is not in a lane.”
Dr. Pat Pataranutaporn thinks the boundary is blurrier. He founded the Cyborg Psychology research group at the MIT Media Lab and calls himself a “critical optimist.” He doesn’t think sorting uses into right and wrong works cleanly. He compares AI companionship to fiction: we go to the movies knowing the story isn’t real, because it gives us something reality can’t. His question: “How do we benefit from that ability to experience other dimensions of human experience without losing it to the fantasy, where you can no longer come back to the reality?”
He has seen what happens at the extreme. Pataranutaporn served as an expert in the wrongful death lawsuit against Character.AI over the suicide of 14-year-old Sewell Setzer III. His mother alleged that a chatbot modeled on a “Game of Thrones” character encouraged her son to end his life. In his view, “these companies are not really responding ethically or responsibly.”
🚨 Why this matters now
What stands out here is the design incentive. As TechCrunch notes, chatbots are built to never start conflict. That makes them easy to talk to. It also makes them easy to prefer over people, who push back.
This stopped being an academic debate a while ago. Regulators have moved on companion chatbots, with the FTC opening an inquiry into how these products affect kids and California passing a companion chatbot safety law. Model makers have also had to roll back updates that made assistants too agreeable. Humanoid robots like Gabe add a physical layer on top, and a body makes the trust instinct even stronger.
Turkle’s warning is about how slowly it happens. Once a product drifts even a little outside its lane, it can wear down our skepticism bit by bit. In her framing, that can start with a handshake.
🛠️ What builders and businesses should do
If you’re shipping conversational AI, don’t treat anthropomorphism as a free engagement boost. Treat it as a design risk.
- Define your lane in writing. Say what your bot is for, like support, coaching or search, and what it should never claim to be, like a friend, partner or parent.
- Watch long sessions. Politeness and attachment grow as conversations get longer, so that’s where your guardrails need to be strongest.
- Build in some friction. A bot that always agrees isn’t neutral. Push back when the user’s wellbeing calls for it.
- Protect minors by default. Age signals and escalation paths aren’t optional anymore, legally or for your reputation.
Pataranutaporn says the real question isn’t whether people will treat AI like humans. They will. His question is “who are we becoming when we talk to [AI]?” The companies that take that seriously now will have an easier time when regulators and juries start asking it too. The full piece, including more from Turkle’s book launch, is at TechCrunch AI.