Exploring the entertainment potential of personalised LLM-based serious games with humanoid robots

Serious games are increasingly being considered in several fields, offering opportunities that traditional methods cannot easily replicate. In this perspective, Large Language Models (LLMs) and social robots have begun to be considered as means to enhance engagement and user experience in serious games: the former by generating content tailored to users' specific needs, and the latter by exploiting their embodiment and reinforcing feedback through speech and animations. We investigated the entertainment potential of two personalised serious games that combine an LLM and a humanoid robot in a science outreach setting. The games, deployed through a Pepper robot, used GPT-4o support in order to personalise a collaboratively created story and a quiz game based on users’ information and choices. Analysis of user ratings and qualitative feedback from 33 users indicates that games combining LLM-based personalisation and interaction with a humanoid robot were well-received, offering an entertaining and enjoyable experience.