Exploring a Virtual Pet to Provide Context Notifications in a Tourism Recommender System: a Pilot Study

Original Abstract

While context-aware personalization has been widely explored in modern tourism Recommender Systems (RS), the delivery of real-time notifications remains a significant design challenge due to issues of intrusiveness and user fatigue. This paper presents a proof-of-concept for a tourism recommendation framework that utilizes a virtual pet as a social mediator for delivering context-aware alerts. The system integrates real-time environmental data - including air quality, noise levels, and weather forecasts - and proximity-based notifications with a Multi-Agent Microservice that generates personalized recommendations based on the user's personality traits and preferences. A within-subjects pilot study (n=11) was conducted to evaluate the feasibility and user acceptance of this pet-mediated approach. Participants interacted with two versions of the system - a baseline without contextual alerts and a version featuring pet-mediated notifications - over a four-week period (two weeks per version) in real-world scenarios. Quantitative and qualitative data were collected to assess engagement, perceived naturalness, notification utility, and acceptance. Preliminary results suggest that the virtual pet effectively can "soften" the perceived intrusiveness of system alerts, making safety-critical information feel more welcome and natural. Furthermore, the character-mediated justifications significantly improved the clarity of the notifications, effectively supporting users in their real-time travel decisions. These findings provide a foundation for using virtual pet companions to enhance the transparency and acceptance of context-aware communication in tourism RS.