GDC 2025 Session Viewer
Amplifying Player Voices with LLMs in Game Development
Joshua Koen (Manager, User Research, Blizzard Entertainment)
Ian Livingston (Director, User Research, Blizzard Entertainment)
Location: Room 2010, West Hall
Date: Thursday, March 20
Time: 1:30 pm - 2:00 pm
Pass Type: All Access Pass, Core Pass - Get your pass now!
Track: Production & Team Leadership
Format: Lecture
Vault Recording: TBD
Audience Level: All
Our players dedicate tens, hundreds, or even thousands of hours to the experiences that we craft for them. Players have valuable feedback about how we can improve those experiences, but it can be challenging for developers to quickly parse vast amounts of feedback and identify ways to improve experiences. The presenters will explore the benefits and risks of using Large Language Models (LLMs) for analyzing player feedback. LLMs offer a way to efficiently analyze feedback at depth and scale that has not been previously possible. However, there are dangers of producing incorrect outcomes that are easy to overlook without a careful approach. Prompt structure, game terminology, and data complexity can easily trip up LLMs. The session will present ways that Blizzard has been validating the use of LLMs for this type of analysis. We will illustrate workflows for using LLMs to enhance processes for analyzing feedback from players.
Takeaway
Attendees will learn how the User Research team at Blizzard incorporates LLMs to efficiently analyze player feedback for game developers. They will gain insights into effective strategies for overcoming challenges and pitfalls and understand the key benefits of LLMs through real-world examples.
Intended Audience
This talk is intended for researchers, insights staff, marketers, and designers who have a vested interest in using player feedback to improve experiences with their video games. Attendees will benefit from general familiarity with user research methods and generative AI technologies, such as ChatGPT.