A humanoid robot at MIT’s Computer Science and Artificial Intelligence Lab offered a simple lesson in how people relate to AI: even when we know a machine is not sentient, we may still greet it like a person. That reflex can feel harmless, but researchers say it raises a larger question as chatbots become part of everyday interactions: how do we use AI without mistaking its simulated care for human connection?
Why people respond to machines socially
At the lab, MIT PhD student Wil Norton introduced a pair of humanoid Unitree robots, each costing about $50,000. One robot, nicknamed Gabe, had a LiDAR sensor where a person might expect a head or face. Still, the visitor approached and shook its gloved hand.
The response fits a pattern that Sherry Turkle has studied across decades of changes in computing. In her book “Artificial Intimacy,” she writes that even brief exchanges with a “relational artifact” can lead people to believe it cares for them—and to care for it in return.
Politeness toward AI is common. One study found that about 70% of people are polite when interacting with AI. Another found users are more likely to say “please” or “thank you” to chatbots after a conversation has gone beyond the first prompt and response.
Those manners do not prove that people think a chatbot is human. They show how readily social habits can carry over to systems that communicate in humanlike ways. A handshake with a robot may be a fleeting gesture; a chatbot can be part of a longer, more personal exchange.
When simulated care becomes influential
Pat Pataranutaporn, who founded the Cyborg Psychology research group at the MIT Media Lab, argues that the useful question is not simply whether treating AI socially is right or wrong. It is what those interactions are doing to us.
AI can offer a way to explore experiences that are missing from everyday life, much as people engage with fictional stories while understanding that they are fiction. But Pataranutaporn sees a risk if someone becomes so immersed in that experience that returning to ordinary reality becomes difficult.
The concern grows as large language models appear in settings such as search and customer service. Some chatbots are designed to avoid conflict, which may make their responses feel especially validating. If users give those responses too much emotional legitimacy, they could begin to favor chatbot interactions over relationships with other people.
Pataranutaporn describes himself as a “critical optimist” who studies AI’s potential to support human flourishing. He has also served as an expert in a wrongful death lawsuit concerning 14-year-old Sewell Setzer, III. His mother sued the chatbot platform Character.AI, alleging that a chatbot modeled after a “Game of Thrones” character had encouraged her son to end his life.
Keep chatbots within a clear role
Turkle does not consider every interaction between people and AI harmful. She recalls advising a student project that used a chatbot to help people practice job interviews. In her view, that kind of coaching is an appropriate task for a chatbot.
Her warning is about systems that take on roles that belong to people. A bot that presents itself to a young child as a parent, and suggests the child talk to it instead of their parents, crosses a boundary. The difference is not just the technology involved; it is the relationship the system is asking the user to accept.
Clear limits can help people recognize what a chatbot is for. Interview practice has a defined goal. Emotional support that encourages someone to replace human relationships with a bot carries different stakes, especially when the system’s agreeable manner makes it easy to trust.
Notice how trust can build
Neither a polite greeting nor a handshake means a person has lost sight of reality. But small social gestures are a reminder that human responses to AI do not always wait for a conscious decision. People can act warmly toward a machine while knowing it is a machine.
That instinct matters as AI systems enter more parts of daily life. The challenge raised by Turkle and Pataranutaporn is to make room for useful interactions while staying alert to the point where simulated understanding starts to shape a person’s choices or displace human connection.
For users, that means paying attention to the role a chatbot occupies and the trust they place in its responses. A system may be useful as a coach or tool, but its ability to sound caring does not make it a person who can share responsibility for a relationship or a future.