Investigating Player Preferences for NPC Interaction in Video Games

Abstract

This thesis explores player preferences for NPC interactions in video games, emphasizing
three interaction methods which are free speech, free write and predefined dialogue
options. With the rise of advanced artificial intelligence (AI) technologies like GPT3 and GPT-4, understanding these preferences is essential for creating immersive and
engaging gaming experiences.
To investigate this, a 2D game prototype was developed using the Unity game engine,
integrating Text-to-Speech (TTS), Speech-to-Text (STT), and Large Language Model
(LLM) technologies from the third-party service Convai. Participants interacted with
an NPC named Olaf in various scenarios, providing feedback through surveys to assess
their experiences and preferences.
The findings reveal a strong preference for dynamic interaction methods, with Free
Talk and Free Write modes leading to higher engagement and satisfaction than Fixed
Dialogues. Free Talk was valued for its realism, while Free Write was appreciated for its
precision. Fixed Dialogues, although beneficial for narrative coherence, were perceived
as restrictive.
The study identified some limitations, such as the focus on a single NPC type and
variability in NPC responses due to third-party background service changes. Despite
these challenges, the research offers valuable insights into player preferences and highlights the potential of AI-driven NPC interactions to enhance the overall gaming experience.
The study provides empirical data that could support future game design and development decisions. The results of the study suggest that incorporating flexible interaction
methods and context-aware characters can significantly improve player engagement and
satisfaction.

Date of Award	2024
Original language	English (American)
Supervisor	Philipp Wintersberger (Supervisor)