AI Dominance Unveiled, ChatGPT-4's Counseling Superiority Stuns, Bing Outsmarts Half of Psychologists in Study

A new study published by Frontiers in Psychology studied similarities between AI and human psychologists and found out that AI is better in understanding human emotions and counseling them. This study was mainly done on ChatGPT-4, Google Bard (also known as Gemini) and Bing and the aim of the study was to assess how well these AI models understand social intelligence. ChatGPT-4 was far better than all the human psychologists that participated in the study, while Bing outperformed half of them. Google Bard was only better than psychologists who were pursuing bachelor's degrees in psychology.

Many Large Language Models (LLMs) are developed in a way that they can easily answer questions, translate languages and even make conversations that are quite similar to humans. This is possible due to structures called neural systems that are responsible for human-like responses from LLMs. Many studies have been done previously that prove that LLMs can help in diagnosing mental health conditions but no study was done about how LLMs can prove to perform with social contexts.

One of the authors of the study, Fahmi Hassan Fadhel, says that since LLMs have proved that they are capable of counseling and psychotherapies, many psychologists can feel threatened as these AI models can easily take away their jobs. The way an AI model can understand human emotions and feelings so well and provide suggestions, it can prove to be more useful than human psychotherapists but it is a very alarming issue.

For this study, 180 male psychologists were included from King Khalid University in Saudi Arabia and were divided according to their educational levels from graduated to doctoral students. The participants and LLMs (ChatGPT, Bard, Bing) were asked to answer 64 scenarios taken from the Social Intelligence Scale. The results were made with two criteria – The soundness of judgment and ability to act wisely in social situations. The results proved that some AI models have advanced in a way that they can even outsmart professional human psychologists.

ChatGPT-4 answered 59 out of 64 questions on the Social Intelligence Scale and surpassed all the human psychologists. The average score of human psychologists was 39.19 for bachelor students and 46.73 for doctoral students. Bing scored 48 out of 64 questions. It outperformed 90% bachelor students and 50% doctoral students. Bing could only answer 40 out of 64 questions. The study shows how quickly AI is developing and excelling in every field. Even though these results by LLMs were impressive, it is also a matter of concern about whether AI should be given authority to talk about sensitive matters such as mental health issues.

Read next: ChatGPT Might Be Kinder and More Cooperative Than Humans, New Study Reveals

Previous Post Next Post