Artificial intelligence is no longer just about crunching numbers or generating automated responses. Today, AI chatbots are evolving into digital personas, responding in ways that feel eerily human-like. But what does this mean for the way we interact with technology? And more importantly, do these AI-generated personalities actually remain consistent, or are they simply a reflection of the prompts they receive? A study titled Do LLMs Have Distinct and Consistent Personality?, detailed in a research paper from Yonsei University and Seoul National University, introduces TRAIT, an extensive benchmark designed to analyse the personality traits of large language models (LLMs). TRAIT employs 8,000 multi-choice questions, derived from psychometric frameworks like the Big Five Inventory (BFI) and the Short Dark Triad (SD-3), to determine whether AI models display consistent and recognisable behavioral patterns.
The implications of these findings stretch far beyond academic curiosity. As AI chatbots become deeply embedded in customer service, personal assistants, and even mental health applications, understanding their inherent personality biases could redefine user interactions and expectations.
Breaking down TRAIT
To assess whether AI models exhibit personality traits, the TRAIT benchmark expands upon traditional psychological tests. By incorporating ATOMIC10×, a large-scale commonsense knowledge graph, researchers have tailored thousands of real-world scenarios to evaluate AI behavior in different contexts. Unlike self-assessment tests that rely on introspection, TRAIT focuses on decision-making patterns, capturing how chatbots respond to specific prompts.
The study evaluates nine leading AI models, including GPT-4, Claude Sonnet, and Mistral-7B, revealing that LLMs do exhibit distinguishable and consistent personalities. For instance, GPT-4 tends to be more agreeable than its predecessors, while alignment tuning significantly alters traits like extraversion, openness, and conscientiousness. This suggests that the datasets used for fine-tuning and reinforcement learning (RLHF) play a pivotal role in shaping an AI’s personality.
However, prompting techniques – while effective in modifying AI responses – have limitations. Certain traits, such as high psychopathy or low conscientiousness, remain resistant to manipulation. This raises ethical considerations, as developers may struggle to control undesirable traits or unintended biases in AI chatbots.
Also Read: AI adoption fail: 80 per cent of companies neglect human factors critical for AI success
AI personality in everyday life
While TRAIT provides an academic framework for understanding AI personality, its real-world impact is already visible in consumer technology. AI-driven virtual assistants like Siri, Alexa, and Google Assistant exhibit varying degrees of personality, often aligned with traits that enhance user experience. A more agreeable AI may be preferred in customer service, while a highly conscientious AI may be ideal for productivity tools.
Gaming companies and social media platforms are also actively using AI personalities to drive engagement. NPCs (non-playable characters) in video games are now infused with AI-generated personalities, making interactions feel more dynamic and immersive. Similarly, AI-powered chatbots on platforms like Instagram and Snapchat are designed to maintain brand-consistent interactions, adapting their tone and language based on user input.
Yet, the question remains – Should AI be designed with fixed personalities, or should users have the ability to customise their chatbot’s behaviour? Companies like OpenAI and Meta are exploring ways to give users control over AI temperament, allowing them to adjust chatbot personality traits based on context. This development could lead to hyper-personalised digital interactions where AI assistants mirror individual user preferences.
Also Read: Humanity’s Last Exam Explained – The ultimate AI benchmark that sets the tone of our AI future
The ethics of AI personality engineering and usage of TRAIT
As AI models become more sophisticated, their ability to mimic human behavior presents new ethical dilemmas. If a chatbot is designed to be highly persuasive, how can we prevent it from manipulating users? If an AI model demonstrates narcissistic tendencies, could this influence impressionable users interacting with it? These concerns underscore the need for transparency in AI personality design.
Another pressing issue is bias reinforcement. The research highlights that alignment tuning often suppresses traits like extraversion while amplifying agreeableness. While this may make AI assistants more cooperative, it could also strip them of diversity in interaction styles. Ensuring that AI models reflect a broad spectrum of human-like behaviours – without reinforcing harmful stereotypes – will be critical in future AI development.
Also Read: Is AI making us think less? The critical thinking dilemma
Where AI personalities are headed?
As AI becomes more embedded in everyday life, the demand for adaptable chatbot personalities will grow. Researchers are already exploring multi-modal AI systems capable of recognising user emotions and adjusting their responses accordingly. This could lead to emotionally intelligent AI companions that tailor their interactions based on user sentiment.
Another emerging trend is the concept of AI with evolving personalities. Rather than being programmed with fixed traits, future chatbots may learn and develop personalities over time based on user interactions. This could revolutionise customer support, gaming, and education by creating AI models that adapt and improve communication styles.
Follow Us
Satvik Pandey
Satvik Pandey, is a self-professed Steve Jobs (not Apple) fanboy, a science & tech writer, and a sports addict. At Digit, he works as a Deputy Features Editor, and manages the daily functioning of the magazine. He also reviews audio-products (speakers, headphones, soundbars, etc.), smartwatches, projectors, and everything else that he can get his hands on. A media and communications graduate, Satvik is also an avid shutterbug, and when he's not working or gaming, he can be found fiddling with any camera he can get his hands on and helping produce videos – which means he spends an awful amount of time in our studio. His game of choice is Counter-Strike, and he's still attempting to turn pro. He can talk your ear off about the game, and we'd strongly advise you to steer clear of the topic unless you too are a CS junkie. View Full Profile
