Designing Adaptive LLM-Based Social Robots for Retail Sales Consultations

Advancements in social robotics and large language models (LLMs) offer opportunities to enhance consumer experiences in retail settings. We investigate how an adaptive LLM-based social robot should be designed to facilitate consumers’ purchase decision-making during sales consultations. Employing a design science research approach and drawing on Social Response Theory and Media Naturalness Theory, we identified key issues and requirements and derived three design principles focusing on the robot’s verbal and non-verbal communication style. We instantiate the design principles in a social robot building on Furhat’s platform and OpenAI LLMs with two variants: rational and emotional. Participants interacted with the social robot in a laboratory experiment receiving product advice. Results indicate that the rational variant significantly increased purchase intentions. Our research contributes to the design of adaptive LLM-based social robots in retail by demonstrating rational verbal and non-verbal communication positively influences consumers' purchase intentions.