Description: With AI-based conversational assistance such as Amazon Rufus becoming increasingly more popular and powerful, they carry the potential to make customers' online experience more personal, engaging and socially rich. What currently mostly remains chat or voice-based interaction, agents such as Rufus could be embodied by multimodal social avatars--with anthropomorphism, i.e., human-like avatars emerging as the dominant theme. To this day, such avatars are directly embedded on the website. This thesis asks the question of how avatars that re-embody in a customer’s local environment e.g., through Augmented Reality, affect users’ perception of and interaction with them.

Goal of Thesis

The goal of this Bachelor/Master thesis (level will determine the scope) aims to study how re-embodied AI assistants across environments (online shop -> personal environment) affect users' perception of them through the following steps:

1. Review of the related literature on conversational assistance / agents / avatars / social cues in online shopping and customer service. -> identify why the environment matters and what it may affect, e.g., privacy?
2. Formulate research hypotheses based on the related work
3. Develop a prototype system with an embodied agent inside an online shop vs. re-embodied in the local space.
4. Evaluate how the two designs affect users across the relevant dimensions identified in the RW.

Requirements

• Strong interest in embodied AI
• Experience with LLMs; ideally, but not required AR/VR/XR or social robots technology
• Familiarize yourself with the concepts of Social Presence, Uncanniness, Engagement, Trust, Competence and Intrusiveness
• You have read the following three papers: [1] [2] [3]

Note: This thesis can also be done with Virtual / Extended Reality or Social Robots for the re-embodiment of the shopping assistant part, depending on the candidates interests.