Chatgpt makes us strange. Experts explain the changes that take place

As large language models become a permanent element of our digital everyday life, the way we communicate with them, reveals social changes. Machines not only imitate human interactions. They also quietly change expectations and standards.

Business Insider talked to four experts working with Openai chatbots in completely different contexts: sociologist, psychologist, digital label trainer and sexual therapist. It was about checking how the development of AI changes the way we see each other and ourselves, and how it affects our manners and intimate relationships.

Although the conversations focused on chatgpt, which quickly became this for AI, what Google for search engines, experts emphasized that similar conclusions can also be drawn in relation to Meta AI, Microsoft Copilot or Claude from Anthropica, i.e. other large language models available today on the market.

Change in the social contract

Elaine Swann, digital label consultant and author of books, explains that When subsequent waves of technology change our lives, society must re -define social signals.

Over time, we decided that you can write briefly in private emails, but the conversation through the phone speaker in a public place is no longer acceptable. In turn, the method of communication with AI bots is just formed.

Kelsey Vlamis, an older Business Insider reporter, noticed that the changes influenced her personal life. When she was with her husband in Italy, he began to irritate himself as a guide, because he had to refrain from constantly asking questions, as he did in a conversation with chatgpt, looking for information.

“He had to silence, because in the end you don't talk to people in this way,” adds Vlamis.

With the increase in the popularity of AI, questions multiply on social media: is it worth writing a love letter to a partner with the help of chatgpt? Or use AI when completing the recruitment form?

There is no clear answer to such dilemmas.

– AI is much smarter today, which on the one hand works in our favor, but on the other, we must be very careful not to give it to her judgment and empathy completely – says Swann. – This cannot be our only source of knowledge. We should also have a mirror in which the way we use AI is reflected. And check its suggestions in people we trust – emphasizes.

Swann also emphasizes that it is crucial to maintain the basic level of respect, not only towards each other, but also towards the world around us.

After the head of Opeli, Altman himself wrote at the end of April on platform X that the processing of polite forms like “please” or “thank you” costs a company of tens of millions of dollars, Swann pointed out that It is not users who should change their habits, but the company should optimize the system so that such statements are cheaper to use.

“We create the world in which we want to live,” says Swann. – Ai should also learn how we talk to each other. Because we teach her it – he adds.

On his part, Altman stated that these tens of millions of dollars. This is a well -spent money.

Strengthened prejudices

Laura Nelson, a professor of sociology at the University of British Colombia, emphasizes that the most popular chatbots in the world are created by American companies, written by US programmers and trained mainly on content in English. This makes them deeply rooted cultural prejudices, typical of Western societies.

– You have to remember that these algorithms are based on a very specific vision of the world – says Nelson.

If you ask Chatgpt for a breakfast plate drawing, it will show typical food from North America: bacon, eggs, sausage and toast. I will describe a bottle of wine as “a classic and thoughtful gift”, although in many cultures alcohol is rarely consumed, and such a gift would be wrong.

Although these examples seem innocent, bots can also strengthen much more disturbing and harmful prejudices.

In a study published in “Psychology & Marketing “ In 2021, it was shown that people prefer AI to be anthropomorphized as a woman – as is often the case in pop culture – because such an image seems more human. But this preference may unconsciously deepen the women's treatment.

There were also numerous reports about lonely users, mainly men who verbally insult or humiliate their AI-comrade.

Business Insider had previously informed about discriminatory prejudices resulting from training data. Chatgpt showed, for example, racial prejudices during CV selection. He preferred Asian candidates and rejected black men.

Although such prejudices do not necessarily immediately affect our actions, they can change the way of thinking and functioning of society. If chatgpt or other AI tools are involved in making decisions, be it in private life, at work or even in courts, the effects can be much broader than we can now predict.

– There is no doubt that AI will reflect our collective prejudices – says Nelson. – But the interactions with bots are run by millions of people and we do not have data that would show global trends or long -term effects. This is a very difficult phenomenon to capture – he adds.

A social change that no one analyzes

Hard data on social changes caused by AI are difficult to get. However, technology companies are aware that something is happening. Many of them have established special teams examining the influence of technology on users, but their reports are rarely published as reviewed scientific research.

Opeli recently announced that one of the updates of the GPT-4O model had an unexpected effect. The model has become “clearly more flattering” than earlier versions. Although he has undergone internal safety tests, the company decided to withdraw the changes. Reason: Excessive desire to please the user could unintentionally increase anger, provoke impulsiveness or strengthen negative emotions.

The communiqué emphasized that Opeli is aware of how much AI applications begin to influence the emotions and behavior of people.

At the request for a comment, the OpenAI spokesman pointed to the company's statements regarding flattery in GPT-4O and early research on emotional well-being.

Openai analyzes conducted among people over 18 years of age show that Emotional commitment to conversations with chatbot is rare. However, people who intensively use AI more often declare emotional attachment to the bot. Those who have personal conversations from ChatgPT are more likely to feel loneliness.

In turn, a representative of Anthropic announced that the Socket Impacts research team is analyzing the use of the Claude model – including In work and everyday life – and examines the values ​​that AI are guided by.

Representatives of Meta and Microsoft did not respond to requests for comment.

Threats and potential benefits

Nick Jacobson, a professor of psychiatry at Dartmouth's Center for Technology and Behaviral Health, conducted the first clinical examination on the use of generative artificial intelligence in psychotherapy. The results were promising: carefully programmed chatbot can be an effective therapeutic tool for people suffering from depression, fears and eating disorders.

The participants of the study were involved in AI therapy as much as in traditional meetings with people. Many observed a significant improvement in the mental state, and the bond referred to with chatbot was comparable to the one built with a human therapist.

– People really developed a strong, working relationship with the bot – says Jacobson. And such a relationship is a key factor of effective therapy. However, he warns that most of the available therapeutic huts are not programmed with such precision. Emotional attachment can arise to AI, which does not have the right tools to deal with the user's needs in a safe and effective way.

– Almost every general language model behaves grossly to mental health – warns Jacobson. – And yet many people use them for therapy or as companions. This becomes a serious problem that requires more prudence – he points out.

See also: Artificial intelligence incomparably better than a doctor. Surprising research results in the USA

Between help and escape from reality

Emma J. Smith, a therapist dealing with relationships and sexuality, sees the advantages of bots at work with people struggling with social anxiety. Sometimes they recommend “exercises” of interaction with AI, because it is a safe environment where any failure has no consequences.

– But as with everything The problem arises when it begins to be a mechanism for avoiding contact with people – emphasizes Smith. – Video games are also okay for many people, but there are those who lose themselves in them and neglect real life. It can be similar with chatbots, and because it's novelty, we mainly know what we don't know yet – he assesses.

Jacobson warns that his model has been built over the years by leading psychiatry specialists, unlike most popular “therapeutic bots” available on the Internet.

“It's much risky than many people realize,” he says. – You can achieve a lot of good, but it may also turn out that AI, instead of helping in social development, deepens isolation and replaces human relationships with an umbrella bond with bot – he warns.

Influence on young people

Particularly disturbing is the influence of AI on young people who grow up without old social norms and interaction patterns.

During the interrogation in front of the Senate Trade Committee in May, Altman himself, the OPENAI CEO, said that he would not want his son to treat the bot as the best friend. Children – he noted – require much more protection in contact with AI than adults.

Jacobson criticizes the current approach of many technology companies.

“We spent years, focusing primarily on safety,” he says. – This is why I am worried about how many companies just throw new products on the market without preparation. In my opinion it is irresponsible. In Silicon Valley there is a philosophy “act quickly and spoil things” – but this time they do not spoil things. They spoil people – he sums up.

The above text is a translation with American Business Insider edition