Ticker

6/recent/ticker-posts

MIT alerts the risks of emotional dependence of Chatgpt

MIT alerts the risks of emotional dependence of Chatgpt

The relationship between humans and conversational artificial intelligence (AI) raises an ever-increasing number of questions. While fiction has already largely taken up this theme – notably with the series Westworld, in which distinguishing humanoids from humans put viewers to the test, or the film Her, dedicated to the love between humans and AI – the subject also challenges researchers.

The Massachusetts Institute of Technology (MIT) has taken an interest in it in its latest study, published on March 21. Researchers from this prestigious institute have thus examined the psychosocial effects of conversational AI like Chat GPT. Their analysis differentiates between interaction modes – neutral voice, engaging voice, and text –, but also between types of exchanges – personal questions, open-ended questions, etc.

Based on the monthly usage of nearly a thousand ChatGPT users and more than 3 million queries, the study also relies on an automated analysis carried out by OpenAI on nearly 40 million messages. "Overall, greater daily usage, regardless of the modes and types of interactions, is correlated with greater loneliness and dependence, but also with an increase in problematic uses and less socialization, the study indicates.

Further research needed

Users who are more emotionally engaged and those who ask the most personal questions face more loneliness. For their part, those who place a high level of trust in AI are more prone to emotional dependence. As for Those with very heavy usage are more likely to consider the chatbot a friend and attribute human emotions to it.

ChatGPT has some 400 million weekly users, and a growing share uses it for personal questions and companionship. According to a YouGov survey, half of Americans between the ages of 18 and 29 feel comfortable discussing their mental health issues with ChatGPT. Its use therefore continues to raise questions.

The MIT researchers conclude: "We emphasize the need for further research to determine whether the ability of chatbots to manage emotional content, without encouraging dependency or replacing human relationships, remains beneficial for general well-being."

Post a Comment

0 Comments