Shooting The Breeze
Shooting the Breeze. A research project developed as part of QUT’s Vacation Research Experience Scheme (VRES), explored how LLM-driven AI agents can shape a player’s sense of presence, connection, and fun. The study examined how chatting with a virtual buddy influenced participants’ perceived experience while engaging with a relaxing, fun, archery-based VR game.
Summary Of Work
Designed and implemented a simple yet engaging VR archery game in Unity (XR-Package) as part of a research project exploring human interaction with LLM-driven AI agents. Implemented core VR mechanics, designed and facilitated observational gameplay testing, and conducted statistical analysis on collected player data. Developed, tested, and refined AI prompts, anti-prompts, and contextual information to shape the agents’ personality and knowledge. Presented research findings in both a scientific paper and formal presentation.
Development Outcomes
Despite the limitations, the project and related study provides some interesting insights into the potential differences in how audiences of differing prior experience in VR and of NPC interactions perceive the social presence and relatability of the ECA.
In the context of a novel gaming experience, where the NPC is only present for pleasant social interactions, participants found the ECA to be a positive social presence which they were highly willing to voluntarily engage with. They also found this entity to be relatable and expressed a desire to have ongoing interactions with it. More experienced players were more likely to find the ECA to have a positive net impact on their experience. This effect did not appear to strongly modify enjoyment regardless of prior experience, but qualitative feedback suggested a high degree of engagement and entertainment. it is unclear however if this is due to the ECA’s superior engagement & responsiveness representing a more enticing source of novelty for experienced participants in comparison to NPCs traditionally found in current games.
These findings support previous conclusions that embodied conversational agents can lead users to treat human-agent interaction like human-human interaction, which can lead to positive effects in terms of engagement, motivation, and interest (Kim et al. 2020).