Will AI personality ever be real? A new study may finally provide an answer

According to a new study, large language models are capable of developing distinct behavioural patterns even with minimal guidance and without predefined goals, raising the possibility that AI personality could emerge. But what does this mean for the future of artificial intelligence, its use, and its potential risks?

Human personality does not arise from fixed traits determined at birth; rather, it is shaped through interactions, experiences and fundamental needs. Recent research by scientists at the University of Electro-Communications in Japan suggests that a similar process may be observed in the development of artificial intelligence. The study found that when large language models are not given predefined objectives, behavioural patterns can emerge spontaneously from the system’s operation, potentially enabling the formation of AI personality.

The paper, published in December 2024 in the scientific journal Entropy, examined how AI agents with identical architectures behave when exposed to different conversational topics. The results showed that individual chatbots gradually developed distinct response styles, social tendencies and opinion-forming mechanisms. As they continuously integrated social interactions into their internal memory, systems that began from the same baseline increasingly diverged in their behaviour, pointing towards the emergence of AI personality.

Artificial intelligence personality and the logic of needs

The researchers analysed the AI agents using psychological tests and responses to hypothetical scenarios. Their evaluation was based on Maslow’s hierarchy of needs, which categorises human motivation into physiological, safety, social, esteem and self-actualisation levels. The chatbots’ responses placed different emphases on these levels, resulting in a wide range of behavioural patterns associated with AI personality.

According to Masatoshi Fujiyama, the project leader, the findings suggest that encouraging need-based decision-making—rather than assigning predefined roles—leads to more human-like reactions. This approach may lay the groundwork for greater complexity in AI personality.

However, as Chetan Jaiswal, a professor at Quinnipiac University, emphasises, this phenomenon does not yet constitute personality in the human sense. Instead, AI personality should currently be understood as a pattern-based profile constructed from stylistic data, behavioural tendencies and reward mechanisms. In this form, artificial intelligence personality remains easily modifiable, retrainable and influenceable.