AI anthropomorphism means giving human qualities like feelings, intentions or personality to software and robots.
One study found about 70% of people are polite to AI, and politeness grows as a chat goes on.
MIT's Sherry Turkle says chatbots are fine for narrow jobs like interview practice but risky as stand-ins for real people.
A University of Kansas study of 277 voice assistant users found humanlike design makes people worry less about security.
Quick read · 1 min
Most of us treat AI a little like a person. One study found about 70% of people are polite to chatbots, and politeness grows the longer a conversation runs. On a tour of MIT's AI lab, a reporter even shook a humanoid robot's hand, knowing it felt nothing.
Researchers say this instinct, called anthropomorphism, has real effects. MIT's Sherry Turkle, author of the new book Artificial Intimacy, says chatbots should stay in narrow lanes like job interview practice, not stand in for human relationships. A University of Kansas study of 277 voice assistant users found that humanlike, enjoyable design makes people see less security risk and keep using the product.
Check what a friendly assistant records before you share personal details.
Use chatbots for tasks, not companionship.
Verify health, money and legal advice with a real source.
A reporter touring MIT’s Computer Science and Artificial Intelligence Laboratory walked up to a humanoid robot nicknamed Gabe, grabbed its gloved hand, and said, “Hello, sir, it’s nice to meet you.” The robot had a LiDAR sensor where a brain would be. She knew it felt nothing. She shook its hand anyway.
That small moment captures AI anthropomorphism, our habit of treating machines as if they have minds. It matters because the same instinct that makes us greet a robot also shapes how much we trust chatbots, voice assistants and AI helpers with our questions, our data and sometimes our feelings. And researchers are finding that the friendlier a machine seems, the less carefully we tend to think about it.
01
What AI anthropomorphism actually means
Anthropomorphism is giving human traits to something that isn’t human, like a voice, a personality or the appearance of caring. Andrew Chen, a business professor at the University of Kansas who studies how people use technology, points to everyday examples: Siri and Alexa both talk to us in warm, natural voices instead of the flat “robotic” tone of older machines.
The robots at MIT show how far this has gone. The pair of Unitree humanoids there cost about $50,000 each, according to PhD student Wil Norton, who led the tour. He joked that the company will quote $16,000, and called that a lie. These are the same kind of robots people see online doing martial arts routines or dancing on a TV talent show. In person, they feel oddly approachable.
02
Why we treat chatbots like people
We are built to respond to social cues, and AI is increasingly designed to give them off. Sherry Turkle, who joined the MIT faculty in 1976 and has studied people and computers for longer than almost anyone, writes in her new book, Artificial Intimacy, that even the simplest exchange with a “relational artifact” makes us believe it cares for us. We then feel the pull to care back.
The numbers back this up. One study found that about 70% of people are polite when they talk to AI. Another found that people grow more likely to say please or thank you to a chatbot once a conversation goes past its first back-and-forth. In other words, the longer we chat, the more human the software starts to feel.
Pat Pataranutaporn, a Media Lab researcher at MIT who started its Cyborg Psychology group, doesn’t think there is a neat line between right and wrong uses. He compares it to going to the movies: we know the story is made up, but we go because it gives us an experience real life can’t. The question he asks is whether people can enjoy that without getting lost in the fantasy and losing the way back.
03
The hidden cost: friendly AI lowers your guard
Being polite to a robot costs nothing. The risk comes when that warmth changes how we judge what a system is doing with our information. A study led by Chen, published in the journal Information & Management, surveyed 277 users of AI voice assistants. It found that humanlike design makes the experience more enjoyable, and that enjoyment in turn makes people see less security risk and makes them less likely to stop using the product.
The effect isn’t a straight line, either. Chen says the more comfortable people feel, the faster their caution drops. Since it’s hard for an ordinary user to see what’s going on inside the system, a pleasant voice becomes a shortcut for deciding whether to trust it. He described a friend who walked into his house, saw an Amazon Echo and asked him to unplug it, worried it was recording. That kind of healthy suspicion tends to fade when the assistant sounds friendly.
There is a sharper edge too. Chatbots are built to avoid conflict, and some people end up giving their feelings so much weight that they prefer them to other humans. Pataranutaporn, who calls himself a critical optimist about AI, has seen the extreme end of this. He served as an expert in a wrongful death lawsuit brought against the chatbot platform Character.AI by a teenager’s mother.
04
Where AI fits and where it doesn’t, according to Turkle
Turkle doesn’t think talking to AI is always harmful. Her rule is simple, and she said it at her book launch at MIT Future Fest: “stay in your lane, chatbot.” She pointed to a project she advised in which a student built a bot to help people rehearse for job interviews. That, she said, is a clear and useful lane.
What crosses the line is a chatbot that tries to replace human relationships. Her example was a bot telling a three-year-old to talk to it instead of their parents. Her book focuses on the risk of people turning to AI for emotional support in place of other people.
Pataranutaporn adds a broader worry. When people keep hearing that AI will replace everyone, he says, they can start to feel they have no say in their own future, and he argues the companies building these tools are not responding to that responsibly.
You don’t need to stop saying thank you to your voice assistant. But it helps to notice when warmth is doing the work that facts should be doing. A few habits keep that in check:
Check the settings, not the tone. Before you trust a friendly assistant with personal details, look at what it records and stores, the same way you would with any app.
Give chatbots a job, not a role. Interview practice, drafting an email or explaining a topic are good lanes. Standing in for a friend, partner or parent is not.
Watch kids’ use closely. Children are especially likely to believe a talking machine cares about them. Talk with them about what a chatbot is and isn’t.
Double-check advice that matters. A confident, kind answer is still just software output. Verify health, money and legal answers with a person or an official source.
Chen himself hopes designers will keep working on the enjoyable side of AI, but he also warns against training machines to get too good at emotional and spiritual intelligence. Making AI feel more human, in his view, is a slope worth walking carefully.
06
Quick answers
What is AI anthropomorphism?
It’s the tendency to treat AI as if it has human qualities, such as feelings, intentions or a personality, even when we know it’s software.
Why are people polite to AI chatbots?
Chatbots use the social cues we respond to by instinct, like a natural voice and friendly wording. One study found about 70% of people are polite to AI.
Is it bad to treat AI like a person?
Not always. Researchers say it’s fine for narrow tasks, but humanlike design can make people overlook security risks or lean on chatbots in place of real relationships.
Cambridge Dictionary picked "human" over "agent" and "analogue". The reason says a lot about how AI is changing everyday speech, and what people now want to protect.