Agenda
What Good Are AI NPCs? Lessons from a Large-Scale Player Study (Presented by NVIDIA)
Thomas Keane (Creative Director, Meaning Machine)
Ben Ackland (Technical Director, Meaning Machine)
Dr. Richard Cole (Lecturer - Digital Futures, University of Bristol)
Dr. Chris Bevan (Lecturer - Immersive Technologies, University of Bristol)
Location: Room 3020, West Hall
Date: Wednesday, March 11
Time: 5:00 pm - 6:00 pm
Pass Type:
Festival Pass, Game Changer Pass -
Get your pass now!
Audience Level:
Intermediate
Track:
Design
Vault Recording: Not Recorded
Audience Level: Intermediate
This session presents findings from a major in-depth player study conducted with the University of Bristol. Featuring over 100 participants, the research provides early academic evidence that AI-driven NPCs - when embedded within carefully authored experiences - can significantly increase player engagement, enjoyment, and creative freedom. We will share the design insights and technical architecture behind these results, including how NVIDIA’s In-Game Inferencing (IGI) SDK integrates with Meaning Machine’s Authored AI systems. The talk ultimately argues for a hybrid future: generative AI guided by strong human authorship, and provides studios with evidence to assess whether AI NPCs are worth pursuing in their own projects.
Takeaway
Finally! Proof that AI NPCs *can* deliver high levels of player enjoyment, immersion, and narrative engagement.
Strong authorial design remains essential; AI-powered experiences with strong authorial control outperform those with weak authorial control.
Authorial control is achieved via combining hand-written content with aggressive LLM wrangling; welcome to the "bully" approach.
Strong authorial design remains essential; AI-powered experiences with strong authorial control outperform those with weak authorial control.
Authorial control is achieved via combining hand-written content with aggressive LLM wrangling; welcome to the "bully" approach.
Intended Audience
This talk is designed for developers and studios who are "on the fence" about the value of generative AI characters & storytelling. It aims to cut through hype and stigma alike, to examine how players actually respond to AI characters - as well as best practices for designing AI-powered experiences that still bear the stamp of human authorship. All backed by real evidence. No pre-requisite knowledge required.