SMU researchers show AI creates realistic game characters

When Jake Klinkert was growing up, his father suggested that since he loved video games, he should consider making them. That simple encouragement left a lasting impression. Klinkert embraced the idea and pursued it academically, ultimately earning a Master of Interactive Technology in Digital Game Development from SMU Guildhall. His early passion for interactive media has since evolved into a cutting-edge research focus. Now a PhD student in the Computer Science Department at SMU’s Lyle School of Engineering, Klinkert is exploring how large language models (LLMs), such as ChatGPT, can be used to create non-playable characters (NPCs) that act and respond more like real people, with consistent personalities and believable emotional responses.

In experiments that generated over 50,500 individual data points, Klinkert and his team found that GPT-4 achieved 73.98% accuracy in maintaining consistent personality traits across a range of interactions. This represents a major improvement over earlier AI models, which scored below 18% in comparable tests. The results suggest that advanced language models are capable of capturing and sustaining distinct personality profiles, a capability that could fundamentally change how game developers approach character creation.

LLMs That Recognize, Understand and Respond to Human Emotions

The research directly addresses a longstanding challenge in the gaming industry: how to create NPCs that display realistic emotional complexity and consistent behavioral patterns. Historically, many games relied on scripted responses that often became repetitive and predictable, breaking player immersion. By contrast, LLM-driven NPCs can adapt their behavior dynamically, providing responses that feel nuanced, context-aware, and emotionally appropriate.

"This represents a shift in how we can approach character development," explained Klinkert. “We've moved from a world where creating believable AI characters required complex systems and extensive technical resources to one where developers can quickly prototype personality-driven characters using text-based interactions.” This accessibility could dramatically lower the barrier to creating more sophisticated, emotionally rich game experiences.

To evaluate the models, the researchers used the International Personality Item Pool questionnaire, a 50-item test derived from the Big Five personality framework, to evaluate three OpenAI models: text-davinci-003, gpt-3.5-turbo-0613, and gpt-4-0613. The Big Five measures personality across five dimensions:

• Openness: High creativity, readily embraces novelty, driven by tackling new challenges, engages in abstract thought
   
• Conscientiousness: How organized and responsible someone is in their work and with other people
   
• Extraversion: How outgoing and social someone is when around other people
   
• Agreeableness: How willing someone is to consider different opinions and work toward shared understanding with others
   
• Neuroticism: How anxiously or emotionally someone tends to view and react to situations
   

Gaming Industry Impact and Applications

Video game companies have long sought to integrate affective computing—technology that can recognize, understand, and respond to human emotions—into interactive experiences. Klinkert’s research demonstrates that advanced language models can enhance these efforts by generating dialogue and decisions that authentically reflect specific personality traits.

The implications of this work extend beyond creating simple dialogue systems. Klinkert envisions NPCs that can retell stories from their unique perspectives, contribute to evolving narratives through improvisation, or solve in-game mysteries guided by their personality-driven intuition. These capabilities could transform gameplay, making virtual worlds feel more immersive, alive, and emotionally engaging.

As players increasingly demand more sophisticated experiences, the ability to create NPCs with believable personalities could become a defining feature for game developers. Imagine a game where characters not only react realistically to the player’s choices but also initiate unique storylines, remember past interactions, and demonstrate consistent behavioral patterns over long periods of play.

SMU researchers show collaboration and future directions 

The study’s complete dataset and results are publicly available through a GitLab repository, enabling other developers and researchers to build upon this foundational work. The team is also exploring partnerships with commercial game development studios to implement these findings in real-world projects, potentially influencing the next generation of narrative-driven and role-playing games.

Klinkert’s collaborators on this research include Corey Clark, deputy director for Research at SMU Guildhall and associate professor of computer science at SMU’s Lyle School of Engineering, and Steph Buongiorno, a recent postdoctoral fellow at SMU Guildhall. Together, they are investigating not only how personality-consistent NPCs can improve player immersion, but also how LLM-driven characters can contribute to more dynamic, interactive storytelling and educational simulations.

Beyond the immediate applications in gaming, this research may also have implications for virtual assistants, educational tools, and social AI systems, where the ability to sustain personality and emotional consistency can improve user engagement and satisfaction. The study illustrates a future where artificial intelligence in interactive environments moves closer to replicating the nuance and depth of human behavior.

By combining expertise in game development, computer science, and AI research, Klinkert and his colleagues are pushing the boundaries of what NPCs can achieve. Their work signals a shift toward more intelligent, responsive, and emotionally compelling virtual characters, opening new opportunities for creators and players alike to explore interactive worlds in ways that were previously impossible.