Sherry Turkle 新书访谈：聊天机器人不该说“我”

从社会科学家到拉响警报的人

Sherry Turkle 在 MIT 研究人与计算机互动已四十年，长期以保持学术距离的社会科学家形象出现。WIRED 记者 Steven Levy 在访谈开头指出，她的新书《Artificial Intimacy: Who We Become When We Talk to Machines》改变了这种写法：它更像一份论战式警告，把过度依赖聊天机器人描述为人的处境危机。Levy 将这本书与《寂静的春天》等唤醒公众的作品相提并论，这是记者的定位，不是研究结论。

Turkle 自己解释了转变的由来。她说写每本书都从好奇出发，但在疫情期间做了大约十次访谈后，中立的观察心态消失了。她描述受访者在家人在受苦、向自己伸出联系的时候，却表示只想和聊天机器人待在一起。她还提到，面对面访谈时自己并不愉快，因为有人不想参与政治，只想活在“聊天机器人乐园”里。这些是访谈中的个案观察，不是抽样调查结果。

为什么盯上第一人称

访谈的核心主张之一，是聊天机器人不该被允许使用“我”和“我”。Turkle 认为，社交媒体已经让人习惯“难过就打开电脑”，而新一代技术“没有理由使用第一人称”。她并不主张所有聊天机器人任何时候都有毒，她明确说自己不愿放弃多年积累的、并非卢德分子式恐慌的可信度，但仍把这种拟人化称为危险的滑坡。

她特别反对用聊天机器人替代付费心理治疗的说法。她提到，自己在严肃访谈中反复听到一种“练习”神话：人们说可以用聊天机器人练习与老板、恋人、老师说话。但她的判断是，当人泡在大语言模型提供的“完全接纳的温汤”里，与机器人对话本身就成了终点，而不是通往真实对话的练习。这里她给出的是论证与观察，不是对照实验的结果。

“人类在环”为何让她反感

Turkle 对职场中的“human in the loop”一词尤其不满。她的解释是，这个说法意味着人已被移出生产流程，只留下做清理工作，比如捕捉幻觉或 AI 没看到的东西。她认为这行不通，因为雇主只会安排少数几个人在环中，这些人基本改变不了什么；他们的部分作用是让公司在被起诉时听起来体面，并给人造成流程没有被交给 AI 的错觉。

这段是她的判断和批评，访谈没有提供具体企业案例或调查数据。它值得作为观点来读：争论的焦点不只是 AI 会不会出错，还包括责任被放在谁身上、那个人是否真有权力改变结果。

陪伴、孤独与不可替代的部分

Levy 提出了一个反方视角：也许强大的聊天机器人是一种新的互动形式，未必比人类对话差，只是不同；有人报告说，机器人给出的正面反馈让他们感到被鼓励、被赋能。Turkle 的回应不是否认这种体验，而是划出边界。她说 AI 永远不会取代人类对话，因为它不怕死、不怕病、不怕痛，不爱自己的孩子，也不在乎你是自杀还是做晚饭。

她进一步说，如果用机器人抚养孩子，或把最私密的对话交给它，我们就会失去联系。但她也没有把孤独者的需求一笔抹掉。面对“很多人没人说话、觉得聊天机器人有帮助”的追问，她的回答是：如果存在孤独危机，那是因为我们在抛弃彼此；给她一个数据中心，她愿意把它变成服务老年人的资源。这是她的规范性主张，不是已被验证的政策效果。

她为什么说现在更有希望

访谈结尾，Levy 问她是否真觉得人与 AI 日益亲密的关系能被阻止。Turkle 说，她比六个月前更有希望，原因是围绕“灭绝事件”、Hugging Face，以及有人误入澳大利亚公务员系统的讨论变多了。她认为人们会愿意同时审视亲密的成本和不亲密的成本。

这段话里的具体指涉在材料中没有展开，读者只能把它理解为她对公共讨论扩大的观察。整篇访谈的立场很清楚：她不是要证明聊天机器人一律有害，而是要求把“亲密”这一层代价纳入对 AI 的公共审视。她的证据主要来自访谈观察与论证，而非新发表的实证研究。

MIT professor Sherry Turkle warns that people have become dangerously enmeshed with chatbots.

During four decades of studying the effects of interaction with computers, Sherry Turkle has developed a reputation as a careful if critical social scientist who approaches her work with academic rigor and distance.

Her new book, Artificial Intimacy: Who We Become When We Talk to Machines, breaks this mold. It is a polemic warning that we have reached a state of emergency where our humanity is at risk from overreliance on soulless machines who pose as our tutors, mentors, friends, colleagues, and lovers. Turkle is aiming for the high bar of books —Silent Spring, The Jungle, Unsafe at Any Speed—that awaken the public to impending crises.

Essentially, this diminutive MIT professor is taking on a multitrillion-dollar industry backed by a government eager to accelerate AI into every corner of our personal and professional lives.

STEVEN LEVY: Your previous books reported results of your research as a social scientist, often leading to troubling conclusions about how people, especially children, interact with computers. But this book reads like you’re breaking the glass and pulling the fire alarm. Why the shift?

SHERRY TURKLE: Like I do with all my books, I went in curious. But after 10 or so interviews, conducted during the pandemic, neutral-curious was no longer my state of mind. This was terrible! Here are these people during the pandemic, their families are suffering, reaching out to them, and they're talking to me about how they really want to just hang with their chatbot. When I was able to interview people face to face, I was not happy. People don't want to do anything in politics because they just want to live in chatbot-land. So the arc of the book ended up as—reality needs you. It couldn't be a worse time to give up on reality.

You write that social media set us up for this.

Social media softened us up by getting us used to the idea that if you're feeling sad or blue, open up your computer. Then it made you mad and kept you with your own kind. But now we have this new technology that has no reason to be first-person.

Do you think that we should ban chatbots from using the first person?

Yes! After a career of having credibility as not being Luddite-crazy, I'm not in the mood to give all that away and say all chatbots are toxic all the time. But this is very dangerous, a slippery slope. This is being used to basically say we don't need to pay for psychotherapy, we can use chatbots. In every serious interview I do, people invoke this myth about practice—that you use chatbots to practice for talking to your boss, your lover, your teacher. But when you’re bathing in the warm soup of total acceptance [from LLMs], talking to chatbots becomes the endgame.

You also have a problem with reliance on chatbots in the workplace, and a particular loathing for the term “human in the loop.”

When they use that term, it means the human has been taken out of the productive process and is there to do some sort of cleanup work, to catch the hallucinations or things the AI doesn’t see. That doesn't work, because employers only have a couple of humans in the loop. Those people basically can't change anything. They are there partly to make it sound good when they're litigated, and partly to give people a false sense that the process hasn't been delegated to an AI. But it has been delegated to an AI.

Maybe powerful chatbots are a new form of interaction that isn’t necessarily worse than human conversation, but different, as people report feeling encouraged and empowered by the positive feedback they provide.

An AI won't ever replace human conversation. It’s really not afraid of death, it’s not afraid of illness, it's not afraid of pain, it doesn't love its children, and it doesn't care if you kill yourself or make dinner. So if we raise our kids with it, or have our most intimate conversations with it, we're going to lose contact.

Still, a lot of people have nobody to talk to. They feel that talking to chatbots helps them.

If there's a crisis of loneliness, it's because we're abandoning each other. Give me one data center, and I will turn it into resources for the elderly.

Studies show that most people are skeptical about AI, and there has been considerable pushback. Yet hundreds of millions of people use them. Even people who push back against data centers are using chatbots.

The goal of this book is to expand the terms of the pushback to include this more intimate connection.

Do you really feel that our growing intimacy with AI can be stopped?

I have more hope now than I did six months ago, because of all of this talk about extinction events and Hugging Face and getting into the Australian civil service by mistake. I feel that people will be willing to look at the intimate as well as the non-intimate costs.