Observations On Artificial Intelligence Applications in Non-Player Characters in Games

Xintong Liu

doi:10.54097/jracrf95

Authors

Xintong Liu School of Computing and Data Science, Xiamen University Malaysia, Sepang, 43900, Malaysia

DOI:

https://doi.org/10.54097/jracrf95

Keywords:

Large Language Model; Deep Learning; Reinforcement Learning; Imitation Learning; Non-player characters.

Abstract

This paper surveys the application of artificial intelligence (AI) in NPC behavior modelling development in video games. The motivation for this work comes from the desire to have more interactive, context-based NPC behavior in video games. It is a well-known fact that NPCs used to behave in predefined ways. But now, with the application of AI technologies such as large language models (LLM), reinforcement learning (RL), and imitation learning (IL), NPCs can now behave in response to the player's behavior and changes in the environment, thereby offering more natural interactions. The paper is organized into three main sections: NPC behavior evolution, NPC and player interaction, and application of AI approaches such as LLM, RL, and IL to make NPC more interactive. The contribution of this paper is to provide an overview of how AI is being applied in NPC behavior modelling in video games, making them more intelligent and emotionally engaging to players, and improving their gaming experience. The paper also addresses some of the challenges associated with current work and possible future works on refining NPC behavior such as hybrid models and multimodal learning.

Downloads

Download data is not yet available.

References

[1] Rivera G, Hullett K, Whitehead J. Enemy NPC design patterns in shooter games. In: Proceedings of the First Workshop on Design Patterns in Games, 2012 May: 1-8.

[2] CD Projekt Red. The Witcher [Video Game]. CD Projekt Red, 2007.

[3] Černý M, Gemrot J, Brom C, Bída M. Using behavior objects to manage complexity in virtual worlds. IEEE Transactions on Computational Intelligence and AI in Games, 2015, 7(4): 355–368. DOI: 10.1109/TCIAIG.2014.2365367.

[4] Ubisoft, NVIDIA. Generative AI and large language models in NPC interactions. Ubisoft, 2024. Available from: https://www.ubisoft.com

[5] CD Projekt Red. The Witcher 3: Wild Hunt [Video Game]. CD Projekt Red, 2015.

[6] Badowski A. Cyberpunk 2077 [Video Game]. 2020.

[7] Inworld AI. Generative AI and NPC interactions in gaming. Inworld AI, 2024. Available from: https://www.inworld.ai

[8] Brown T, Mann B, Ryder N, Subbiah M, Kaplan J D, Dhariwal P, et al. Language models are few-shot learners. Advances in Neural Information Processing Systems, 2020, 33: 1877–1901.

[9] Aslan P. Understanding LLMs for game NPCs: an exploration of opportunities. 2024.

[10] Everstone Studios. Where Winds Meet [Video Game]. Everstone Studios, 2024.

[11] sSong L. LLM-driven NPCs: cross-platform dialogue system for games and social platforms. arXiv, 2025. Available from: https://arxiv.org/abs/2504.13928

[12] Senanayake C, Campus H. Dynamic NPC AI using reinforcement learning for an enhanced gaming experience. 2025.

[13] Korkiakoski M, Sheikhi S, Nyman J, Saariniemi J, Tapio K, Kostakos P. An empirical evaluation of AI-powered non-player characters' perceived realism and performance in virtual reality environments. arXiv preprint, 2025. arXiv:2507.10469.

[14] Ng A Y, Russell S. Algorithms for inverse reinforcement learning. In: ICML, 2000 Jun, Vol. 1, No. 2: 2.

[15] Rockstar Games. Red Dead Redemption 2 [Video Game]. Rockstar Games, 2018.