Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
102 tokens/sec
GPT-4o
59 tokens/sec
Gemini 2.5 Pro Pro
43 tokens/sec
o3 Pro
6 tokens/sec
GPT-4.1 Pro
50 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

Player-Driven Emergence in LLM-Driven Game Narrative (2404.17027v3)

Published 25 Apr 2024 in cs.CL and cs.AI

Abstract: We explore how interaction with LLMs can give rise to emergent behaviors, empowering players to participate in the evolution of game narratives. Our testbed is a text-adventure game in which players attempt to solve a mystery under a fixed narrative premise, but can freely interact with non-player characters generated by GPT-4, a LLM. We recruit 28 gamers to play the game and use GPT-4 to automatically convert the game logs into a node-graph representing the narrative in the player's gameplay. We find that through their interactions with the non-deterministic behavior of the LLM, players are able to discover interesting new emergent nodes that were not a part of the original narrative but have potential for being fun and engaging. Players that created the most emergent nodes tended to be those that often enjoy games that facilitate discovery, exploration and experimentation.

PDF HTML Abstract

Player-Driven Emergence in LLM-Driven Game Narrative

The paper, "Player-Driven Emergence in LLM-Driven Game Narrative," explores the interplay between LLMs and player agency in the context of text-based games, focusing on how LLMs can promote emergent gameplay and narratives. This work specifically investigates the implementation of GPT-4 within a text-adventure game called "Dejaboom!" to assess the model's role in facilitating novel narrative pathways through player interactions.

Core Concepts and Methodology

The research revolves around integrating LLMs to dynamically generate non-player character (NPC) dialogue in real-time, providing players with an unprecedented level of interaction flexibility. This dynamic interaction aims to transcend traditional predefined game dialogues, thus offering a richer set of narrative possibilities. Through engagement with GPT-4, players can diverge from the fixed narrative tree, thereby introducing emergent elements that enhance gameplay experience.

A user paper was conducted with 28 participants, highlighting how players' interactions with the model led to creative deviations from the original narrative. By converting gameplay logs into a narrative graph, the paper identifies "emergent nodes"—unexpected narrative elements introduced by players as they explore new strategies within the game environment.

Findings

The analysis revealed various categories of emergent nodes introduced, such as:

  • Extracting Information from NPCs: Players employed novel strategies to garner information from NPCs, such as deceit or manipulation, which were not originally anticipated by the game designers.
  • Suggestions for Additional Assets: Players sought enhancements in the game world, implicitly suggesting the inclusion of new objects or NPCs.
  • Creative Problem Solving: Players devised inventive ways to uncover hidden information or achieve game objectives, reflecting their problem-solving approaches.

These observations underline the potential for player-driven emergence, facilitated by the LLM's non-deterministic nature, to enrich player experience and offer valuable insights to game designers for narrative expansion.

Implications and Future Directions

The implications of this research are multi-faceted. Practically, it provides a blueprint for integrating LLMs in game narrative design, empowering designers to adopt a more iterative, player-centered approach. Theoretically, it suggests a framework for understanding how AI can augment creative processes in gaming, expanding the sphere of narrative possibilities.

From a development perspective, the paper indicates that incorporating LLM-driven dialogue can foster a more engaging and exploratory gaming environment. The categorization of emergent nodes offers designers insights into player behavior, which could influence future game mechanics and narrative design. Additionally, this player-driven feedback can be integral in refining the interactive elements of games, leading to more sophisticated AI-driven experiences.

Moreover, by correlating player motivation with the tendency to create emergent nodes, the paper hints at a personalized gaming experience where AI tailors narrative complexity to individual player profiles. Future explorations could investigate this personalization further, optimizing AI responses based on player interaction patterns.

Conclusion

The paper successfully illustrates the potential of LLMs for facilitating emergent gameplay through player-NPC interactions. By analyzing the emergent pathways introduced by players, designers can gain insights into crafting more dynamic and inclusive game worlds. This work highlights the evolving role of AI in interactive entertainment, paving the way for future innovations in game design that leverage the synergy between human creativity and AI capabilities.

File Document Download Save Streamline Icon: https://streamlinehq.com PDF File Document Download Save Streamline Icon: https://streamlinehq.com Markdown Bookmark Chat Bubble Oval Streamline Icon: https://streamlinehq.com Chat (Pro)
Definition Search Book Streamline Icon: https://streamlinehq.com
References (29)
  1. J. van Stegeren and J. Myśliwiec, “Fine-tuning gpt-2 on annotated rpg quests for npc dialogue generation,” in Proceedings of the 16th International Conference on the Foundations of Digital Games, 2021, pp. 1–8.
  2. S. Värtinen, P. Hämäläinen, and C. Guckelsberger, “Generating role-playing game quests with gpt language models,” IEEE Transactions on Games, 2022.
  3. Q. C. Gao and A. Emami, “The turing quest: Can transformers make good npcs?” in Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 4: Student Research Workshop), 2023, pp. 93–103.
  4. V. Kumaran, J. Rowe, B. Mott, and J. Lester, “Scenecraft: automating interactive narrative scene generation in digital games with large language models,” in Proceedings of the AAAI Conference on Artificial Intelligence and Interactive Digital Entertainment, vol. 19, no. 1, 2023, pp. 86–96.
  5. M. O. Riedl and A. Zook, “Ai for game production,” in 2013 IEEE Conference on Computational Inteligence in Games (CIG), 2013, pp. 1–8.
  6. R. Volum, S. Rao, M. Xu, G. DesGarennes, C. Brockett, B. Van Durme, O. Deng, A. Malhotra, and W. B. Dolan, “Craft an iron sword: Dynamically generating interactive game characters by prompting large language models tuned on code,” in Proceedings of the 3rd Wordplay: When Language Meets Games Workshop (Wordplay 2022), 2022, pp. 25–43.
  7. L. Huang and X. Sun, “Create ice cream: Real-time creative element synthesis framework based on gpt3.0,” in 2023 IEEE Conference on Games (CoG), 2023, pp. 1–4.
  8. S. Chauvin, G. Levieux, J.-Y. Donnart, and S. Natkin, “Making sense of emergent narratives: An architecture supporting player-triggered narrative processes,” in 2015 IEEE Conference on Computational Intelligence and Games (CIG), 2015, pp. 91–98.
  9. D. Ippolito, A. Yuan, A. Coenen, and S. Burnam, “Creative writing with an ai-powered writing assistant: Perspectives from professional writers,” arXiv preprint arXiv:2211.05030, 2022.
  10. G. Franceschelli and M. Musolesi, “On the creativity of large language models,” arXiv preprint arXiv:2304.00008, 2023.
  11. M.-A. Côté, A. Kádár, X. Yuan, B. Kybartas, T. Barnes, E. Fine, J. Moore, M. Hausknecht, L. El Asri, M. Adada et al., “Textworld: A learning environment for text-based games,” in Computer Games: 7th Workshop, CGW 2018, Held in Conjunction with the 27th International Conference on Artificial Intelligence, IJCAI 2018, Stockholm, Sweden, July 13, 2018, Revised Selected Papers 7.   Springer, 2019, pp. 41–75.
  12. OpenAI, “Gpt-4 technical report,” ArXiv, vol. abs/2303.08774, 2023. [Online]. Available: https://api.semanticscholar.org/CorpusID:257532815
  13. J. Urbanek, A. Fan, S. Karamcheti, S. Jain, S. Humeau, E. Dinan, T. Rocktäschel, D. Kiela, A. Szlam, and J. Weston, “Learning to speak and act in a fantasy text adventure game,” in Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), 2019, pp. 673–683.
  14. M. Hausknecht, P. Ammanabrolu, M.-A. Côté, and X. Yuan, “Interactive fiction games: A colossal adventure,” in Proceedings of the AAAI Conference on Artificial Intelligence, vol. 34, no. 05, 2020, pp. 7903–7910.
  15. S. Yao, R. Rao, M. Hausknecht, and K. Narasimhan, “Keep CALM and explore: Language models for action generation in text-based games,” in Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), B. Webber, T. Cohn, Y. He, and Y. Liu, Eds.   Online: Association for Computational Linguistics, Nov. 2020, pp. 8736–8754. [Online]. Available: https://aclanthology.org/2020.emnlp-main.704
  16. Y. Xu, M. Fang, L. Chen, Y. Du, and C. Zhang, “Generalization in text-based games via hierarchical reinforcement learning,” in Findings of the Association for Computational Linguistics: EMNLP 2021, 2021, pp. 1343–1353.
  17. Y. Xu, M. Fang, L. Chen, Y. Du, J. Zhou, and C. Zhang, “Perceiving the world: Question-guided reinforcement learning for text-based games,” in Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), 2022, pp. 538–560.
  18. W. Zhou, X. Peng, and M. Riedl, “Dialogue shaping: Empowering agents through npc interaction,” arXiv preprint arXiv:2307.15833, 2023.
  19. Q. Tan, A. Kazemi, and R. Mihalcea, “Text-based games as a challenging benchmark for large language models,” 2023.
  20. C. F. Tsai, X. Zhou, S. S. Liu, J. Li, M. Yu, and H. Mei, “Can large language models play text games well? current state-of-the-art and open questions,” arXiv preprint arXiv:2304.02868, 2023.
  21. S. Basavatia, S. Ratnakar, and K. Murugesan, “Complexworld: A large language model-based interactive fiction learning environment for text-based reinforcement learning agents,” in International Joint Conference on Artificial Intelligence, 2023.
  22. G. Wang, Y. Xie, Y. Jiang, A. Mandlekar, C. Xiao, Y. Zhu, L. Fan, and A. Anandkumar, “Voyager: An open-ended embodied agent with large language models,” arXiv preprint arXiv:2305.16291, 2023.
  23. A. Kalbiyev, “Affective dialogue generation for video games,” Master’s thesis, University of Twente, 2022.
  24. M. U. Nasir and J. Togelius, “Practical pcg through large language models,” arXiv preprint arXiv:2305.18243, 2023.
  25. G. Todd, S. Earle, M. U. Nasir, M. C. Green, and J. Togelius, “Level generation through large language models,” in Proceedings of the 18th International Conference on the Foundations of Digital Games, 2023, pp. 1–8.
  26. S. Värtinen, P. Hämäläinen, and C. Guckelsberger, “Generating role-playing game quests with gpt language models,” IEEE Transactions on Games, pp. 1–12, 2022.
  27. K. Frans, “Ai charades: Language models as interactive game environments,” in 2021 IEEE Conference on Games (CoG), 2021, pp. 1–2.
  28. Y. Xu, S. Wang, P. Li, F. Luo, X. Wang, W. Liu, and Y. Liu, “Exploring large language models for communication games: An empirical study on werewolf,” arXiv preprint arXiv:2309.04658, 2023.
  29. E. Akata, L. Schulz, J. Coda-Forno, S. J. Oh, M. Bethge, and E. Schulz, “Playing repeated games with large language models,” arXiv preprint arXiv:2305.16867, 2023.
User Edit Pencil Streamline Icon: https://streamlinehq.com
Authors (13)
  1. Xiangyu Peng (33 papers)
  2. Jessica Quaye (6 papers)
  3. Weijia Xu (23 papers)
  4. Chris Brockett (37 papers)
  5. Bill Dolan (45 papers)
  6. Nebojsa Jojic (43 papers)
  7. Gabriel DesGarennes (2 papers)
  8. Ken Lobb (2 papers)
  9. Michael Xu (18 papers)
  10. Jorge Leandro (3 papers)
  11. Claire Jin (4 papers)
  12. Sudha Rao (23 papers)
  13. Portia Botchway (3 papers)
Citations (7)
View on Semantic Scholar
Youtube Logo Streamline Icon: https://streamlinehq.com

YouTube

HackerNews

  1. Player-Driven Emergence in LLM-Driven Game Narrative (118 points, 122 comments)
Reddit Logo Streamline Icon: https://streamlinehq.com

Reddit

  1. Player-Driven Emergence in LLM-Driven Game Narrative (3 points, 2 comments)
  2. Player-Driven Emergence in LLM-Driven Game Narrative (1 point, 1 comment)