AI chatbots have been pushing the boundaries of artificial intelligence for many years. Nowadays, AI chatbots are integrated into almost every industry there is, operating as a streamlining tool that improves operations and increases productivity. The constant development of new features continuously transforms them to be more than just communication tools.
In a new study conducted by researchers at Stanford School of Humanities and Sciences, the latest version of ChatGPT underwent rigorous examination to determine the extent of its similarity to the average conversational aptitude of human participants. The result was astounding, and it proved just the capability of AI chatbots to mimic human personalities by learning from previously inputted data.
New Study Finds That AI Chatbots Shows Human-like Personality Traits
Image from Quora
AI chatbots have revolutionized many industries due to their various capabilities, but their natural language processing feature is one of the most transformative. AI chatbots are employed in various fields such as education, business, and marketing. Its ability to engage in meaningful and productive conversations allows it to be effective in e-commerce, government services, and customer service.
Researchers at the Stanford School of Humanities and Sciences put Open AI’s Chat Generative Pre-trained Transformer (GPT) series to the test to measure the chatbot's human-like capabilities. The study, led by University of Michigan computer scientist Qiaozhu Mei, tested previously released ChatGPT-3.5 and the newly released ChatGPT-4 to see just how much similarities they have with the conversational aptitudes of 108,000 individuals from over 50 countries who participated in the research.
In the study, both ChatGPT-3.5 and ChatGPT-4 were trialed with the OCEAN Big Five personality questionnaire. The questionnaire is used to determine a person’s personality profile by measuring their openness to experience (O), conscientiousness (C), extraversion (E), agreeableness (A), and neuroticism (N).
Apart from the comprehensive, science-backed OCEAN Big Five test, both OpenAI’s chatbots also underwent the MobLab Classroom economics experiment platform which is commonly used to gauge student’s economic principles and strategies when placed in a controlled environment. Both AI chatbots were directly involved in games such as the Dictator Game, Trust Game, Bomb Risk Game, Ultimatum Game, Public Goods Game, and the Prisoners Dilemma Game—all of which were replayed 30 times to ensure that the data gathered was consistent.
AI Chatbots With Human-like Personality and Decision-Making Skills
Image from Medium
Upon extraction of the results, the researchers discovered that both ChatGPT-3.5 and ChatGPT-4 showed notable similarity in scores as compared to the scores of human participants. The similarities extended from the results of the OCEAN Big Five personality test, with just a slight difference between the previous and the latest OpenAI Chatbot series release.
ChatGPT-4 showed similar results with human participants in all personality traits. ChatGPT tested the same, but with a slight difference when it comes to openness to experience (O).
According to Mei, this study was significant since it proved that both OpenAI’s chatbots succeeded the Turing Test, which is used to measure a machine's ability to exhibit intelligent behavior compared to a human. The results of the study proved that both OpenAI chatbots can create more than just human-like speech, but human-like actions and personalities as well.
“By bringing classic economic games into a Turing test, we, for the first time, could profile AI behavior through their actions, not just their words,” said Mei.