人们对 AI 助手说话习惯粗暴直接 | Psychology Today

正文字数894 word

【Para. 1】Most of us would probably say we’re reasonably polite, but there’s one exception: We are rude to our AI assistants. When we speak or type instructions to them, most of us don't include “please” or “thank you,” and we don’t use indirect requests that are a common form of politeness in English, like “Could you send a message?” Instead, we use direct commands like “Send a message.” “Make an itinerary.” “Read this instruction file. Document all your work. Do not skip steps.”

【Para. 2】This is not how we talk to each other, and if someone talked this way to their coworkers, they’d get voted Jerk of the Month.

【Para. 3】The Importance of Politeness

We dislike rude people because they’re cheaters, skipping the effort to be polite while the rest of us are putting in the work. Making some effort to be polite is essential to our functioning in society, where we’re constantly needing to interact with people outside our core family and friend group. Being polite to these strangers signals that we will cooperate with them and that we merit their cooperation in return: holding the elevator, moving aside to let someone pass, calling an ambulance when someone is hurt. Politeness sustains the cooperation that holds society together.

【Para. 4】Being Rude to AI

If politeness is so essential, then why skip it when talking to AI? Because we don’t need it for AI, and we may even be better off without it. The act of talking is cognitively demanding, and omitting the words and phrases that signal politeness could make talking easier. AI assistants are programmed to cooperate no matter how we ask, and sometimes being rude to them can actually yield better performance, compared to polite requests.

【Para. 5】How Talking to AI Could Affect our Human Interactions

Despite these apparent upsides of being rude to AI assistants, we should also consider the potential downsides. I study talking, and one likely downside is that all this rude talking to AI will lead us to speak more rudely to each other. As with almost everything AI these days, it’s too soon to tell the scope of this problem. Nonetheless, the science of talking can let us make educated guesses about whether AI will undermine politeness in our society. The logic of this argument has three steps:

【Para. 6】1. Our Past Talking Shapes Our Future Talking. When we turn our ideas into speech or writing, our brains tend to reuse words and phrases that we’ve often said before. Reusing our favorite words and talking styles gives us well-practiced go-to expressions that we can access quickly, making talking more efficient. This recycling is so pervasive that it’s plausible that the rude talking style for AI will get reused in future conversations with each other, kicking politeness to the curb.

【Para. 7】2. Talking Styles Spread Across Groups. You might think I'm overly worried here; surely we can tell the difference between commands to AI assistants and talking to our friends? Well, not entirely. We can keep talking styles for different groups separate to some degree—for example, saying “tummy” to a toddler but “stomach” to adults. But the separation isn’t complete, and people can start reusing a common talking style with another group in awkward ways. For example, when college students go back home for a visit, they often unconsciously reuse their now-familiar college talking style on their hometown families and friends, sometimes creating hurt feelings when they sound "different" or "snooty". Examples of talking styles spilling over from one group to another suggest that our rude AI talk really could get reused during our human conversations.

【Para. 8】3. AI Chatbot Companions Also Make Us Rude. Users of AI chatbots for companionship don’t typically issue commands and aren’t rude to the chatbots, but those interactions create problems for politeness too. Chatbot companions, developed by companies such as Replika, Character AI, and others, encourage users to be emotionally engaged with the chatbot and to talk endlessly about themselves. Users then start reusing this egotistical talking style, including in their conversations with other humans. This self-centered conversation style is a clear violation of social norms for cooperating and letting everyone have a turn in conversations. Chatbot users even report that they’ve become less interested in talking with humans; they’d rather be egotistical and talk about themselves to a chatbot instead of having to cooperate in back-and-forth conversation with humans.

【Para. 9】Politeness is a big deal

With AI being both praised and blamed for job loss, brain rot, and everything else these days, should we bother adding politeness to the pile of worries? Yes, because of the essential role politeness has in keeping us cooperating and keeping our society functioning. Politeness already faces many other assaults—political polarization, a reduction in face-to-face talking, inequality, and more. The rudeness-inducing effects of talking to AI may push us further away from the civility and cooperation that every society needs.

【声明】：本文原文摘选自Psychology Today，原文版权归杂志所有，仅供个人学习交流使用。