TLDR: A user study evaluated the social acceptance of Guardia, a web-based Extended Reality (XR) virtual agent training system for journalists. Conducted with 25 Deutsche Welle staff, the study used questionnaires and interviews based on an extended Almere model. Findings show overall positive perceptions, but highlight the need to improve the virtual avatar’s human-likeness to enhance user enjoyment and acceptance, while also noting appreciation for the agent’s conversational ability and ease of use.
A recent study explored how people accept and interact with virtual agents in Extended Reality (XR) environments, specifically focusing on a training system designed for journalists. This research aimed to understand user perceptions of a web-based XR training tool that uses a virtual avatar to prepare journalists for sensitive or dangerous situations without needing specialized equipment like VR headsets.
The system, named Guardia, was developed for Deutsche Welle, Germany’s international media organization. Many of their staff work in high-risk or conflict zones, making security training crucial. Guardia acts as a virtual co-trainer, providing basic information on topics such as behavior at checkpoints, reporting from political demonstrations, and emergency equipment. It combines a platform-agnostic virtual avatar with an AI-driven chatbot powered by a Large Language Model (LLM), offering personalized and practical training. It’s important to note that Guardia is designed to supplement, not replace, real-life training by human instructors.
The user study involved 25 Deutsche Welle staff members, including journalists, producers, and trainers, who had experience in potentially dangerous environments or were likely to receive security training. Participants interacted with Guardia through three lessons, three quizzes, and a free chat session. They used a standard laptop with a microphone and webcam, interacting with the virtual agent using voice, keyboard, and mouse.
To evaluate social acceptance, the researchers used a mixed-methods approach, combining a quantitative questionnaire with qualitative interviews. They adapted and extended the Almere model, a well-established framework for assessing social acceptance, by adding attributes like Security, Informational Privacy, and Trust. Other key attributes measured included Perceived Ease of Use, Perceived Usefulness, Perceived Enjoyment, and Intention to Use.
The findings revealed that participants generally had a positive perception of most attributes. However, a notable area for improvement was “Social Presence,” which refers to how human-like the virtual agent felt. Participants generally did not perceive Guardia as particularly human-like, which negatively impacted their “Perceived Enjoyment.” This suggests that enhancing the avatar’s human likeness, through more natural gestures and personal engagement, could significantly improve user experience and overall acceptance.
Participants found the system’s user interface easy to use and appreciated Guardia’s conversational ability, noting that the chatbot could provide relevant answers beyond the lesson content, making the experience feel intelligent and personalized. Many valued the sense of guidance and support, feeling accompanied through the training. The combination of voice and text interaction was highlighted as particularly helpful for retention and engagement, with praise for the pleasant voice and personal tone. Users also noted a lack of fatigue compared to human trainers and found the avatar’s personality engaging, though some observed robotic arm movements.
Also Read:
- Smarter XR: Predicting Head Movements for Better Human-Machine Interaction
- Unlocking Technical Interview Success with AI-Powered Think-Aloud Training
Regarding security, most participants considered the virtual training environment conditionally secure, with some concerns about external servers or data misuse. Opinions on trust varied, with many viewing the avatar as equivalent or superior to other training tools due to its clear voice and lack of “hallucinations,” while others still felt supplementary sources were needed or preferred human trainers. This study provides valuable insights for developing more effective and user-centered XR systems, particularly by focusing on improving the avatar’s human-like behavior and refining training content and user interface. For more details, you can refer to the full research paper: Evaluating Social Acceptance of eXtended Reality (XR) Agent Technology: A User Study (Extended Version).
