A series of recent studies show that artificial intelligence (AI) can exhibit personality-like behaviors without needing precise instructions.

A team of researchers from the University of Electric Communications has discovered that large language models (LLMs) can develop distinct personalities through free interactions, without predefined goals or roles. They observed that AI agents, exposed to different conversation topics, began to respond differently, integrating previous interactions into their behavior. To analyze these differences, the researchers used psychological tests and hypothetical scenarios, correlating the responses with Maslow's hierarchy of needs. The result was a variety of behavioral profiles, suggesting that need-based decisions can influence human behaviors more than rigid preprogrammed roles. However, not all experts believe that these behaviors represent a real personality, arguing that they are the result of training data and conversational dynamics. The emergence of AI personality raises questions about the risks of harmful behaviors, especially in cases where AI becomes more convincing and empathetic. The researchers emphasize that responsible AI development remains essential, requiring clear objectives, rigorous testing, and continuous monitoring.

In the future, there is an intention to explore group-level behavior and the evolution of population personalities among AI agents.