Artificial intelligence (AI) is revolutionizing human interaction. With the development of immersive technologies like virtual reality (VR), researchers are discovering new ways that AI can influence human behavior. One relevant area of research is artificial social influence, where AI agents—particularly those embodied in virtual environments—interact with humans in meaningful and persuasive ways.
A recent study from Michigan State University, led by researchers Sue Lim, Ralf Schmälzle, and Gary Bente, has explored the potential of AI-driven virtual agents to engage humans in natural health-related conversations in a VR setting. This research was made possible by leveraging the WorldViz Vizard VR platform, which enabled the team to create realistic, immersive virtual environments where participants interacted with AI-embodied conversational agents (ECAs).
Conceptual figure illustrating the study design and research question/hypotheses.
By combining VR with advanced AI, researchers can simulate human-like interactions with embodied AI agents, or VR-ECAs. The WorldViz Vizard platform provided the infrastructure for the study, offering real-time, interactive 3D environments that are crucial for studying complex social behaviors in controlled settings.
In this study, VR-ECAs were designed to engage participants in health-related conversations, using cutting-edge AI language models for natural, real-time dialogue. Unlike traditional text-based conversational agents (CAs), the VR-ECAs had a physical presence within the virtual environment, displaying non-verbal cues such as eye contact and lip synchronization, further enhancing the realism and influence of the interaction.
One of the key aspects of this study was the investigation of similarity—in particular, how matching the gender of the virtual agent to that of the participant could influence the outcomes of the interaction. The researchers found that when participants interacted with a gender-matched AI agent, their gaze duration and overall attention increased. Interestingly, this effect was stronger for male participants, suggesting that the perception of similarity plays a crucial role in enhancing the engagement and influence of AI agents.
Female participants, regardless of the gender of the AI agent, tended to like the agents more, which highlights potential gender-based differences in how AI is perceived. These findings underscore the importance of designing AI agents that cater to individual user characteristics, such as gender, to maximize their persuasive impact.
By offering a seamless integration of AI technologies, including text-to-speech and speech-to-text systems, the WorldViz Vizard VR software platform allowed the research team to create lifelike virtual agents capable of mimicking human behaviors in real-time. Moreover, the SightLab VR Pro plugin for Vizard allows for advanced features, such as gaze tracking and real-time movement, providing researchers with the ability to capture detailed biobehavioral data, such as eye-tracking during interactions, which was key to measuring the effects of artificial social influence. Additionally, SightLab VR Pro now offers any researcher the ability to easily add intelligent AI agents to experiments.
Vizard’s capabilities go beyond just visualization; they allow researchers to simulate authentic social environments, making it possible to study how humans interact with AI in ways that feel natural and spontaneous. This is particularly important in studies of artificial social influence, where the authenticity of the interaction can significantly impact the results.
This groundbreaking research not only highlights the potential of AI and VR technologies to influence human behavior but also opens new avenues for further exploration. The combination of AI-driven virtual agents with immersive VR environments presents a unique opportunity to study social influence in a controlled yet lifelike setting. With platforms like WorldViz Vizard, researchers can delve deeper into the nuances of human-AI interaction, exploring factors like similarity, non-verbal communication, and behavioral outcomes in unprecedented detail.
As AI continues to evolve, the integration of these technologies into VR platforms will likely expand, leading to more sophisticated and human-like virtual agents capable of influencing behavior in areas such as health coaching, education, and even marketing.
In summary, the study from Michigan State University demonstrates the power of combining AI and VR to explore the effects of artificial social influence. With the WorldViz Vizard platform, researchers were able to create realistic, immersive environments that enabled them to investigate how AI agents can shape human behavior. As this field of research grows, we can expect to see even more innovative uses of VR and AI in creating socially intelligent systems that not only interact with but also influence us in ways that mirror real-world human interactions.
Read more about how you can add intelligent AI agents to your experiment using the SightLab VR Pro plugin for Vizard in this blog post.
Please contact us at sales@worldviz.com to discuss your academic research interests.
