Embodied AI and Social Robotics in Assisted Shopping Interaction

In this interdisciplinary project we develop and evaluate ‘negotiation processes’ with an autonomous shopping robot (no WoZ). We are interested in designing collaboration and especially non-collaboration in decision-making processes in shopping interaction between robot and customer. We chose shopping as a field of study as it is a popular and traditional field of AI (e.g. amazon) and robotics (Yamazaki et al., 2022) . As shopping often includes several decision-making processes (e.g. during a grocery shopping trip each item requires a decision-making), it allows for systematic studying of “doing assistance” in this decision-making processes. Our focus is directed to non-collaboration for two reasons. A) AI behind shopping algorithms is normally based on economic principles encouraging the customer to buy the product or, better even, buy more products. However, in my empirical data on assistive shopping we could see that assistants often aim to discourage a person of buying something (Krummheuer, 2020, 2021). B) Rejections and other forms of ‘non assistance’ are interactionally seen not preferred actions and of high value as they make underling social norms and expectations more easily accessible in analysis. Therefore we aim to transfer this knowledge gained from a multimodal interaction analysis to the development of robots. We will design and conduct an experiment with a shopping robot that performs different negotiation strategies.

The project is funded by the Humanistic AI Cluster at the Department for Communication and Psychology, Aalborg University.