The evolving landscape of consumer transactions, particularly the widespread adoption of automated self-checkout terminals and self-order kiosks, presents an effective approach for shopping everyday items, or low involvement products, such as groceries. Nevertheless, when dealing with high involvement products, such as smartphones, consumers display a preference for personalized guidance and assistance offered by human salespersons. The advent of Generative AI, specifically language models like ChatGPT, presents potential avenues for enhancing the automated shopping experience by integrating human-like and tailored interaction capabilities. The central question of the project is: How can we design a robot shopping assistant, utilizing large language models, to offer guidance to consumers for making significant product purchases?

Goal of the thesis

The primary objective is to develop an application that facilitates human-like sales dialogues, leveraging a blend of ChatGPT and the social robot, "Furhat" (Video). Both essential technologies are supplied by the institute, and an initial prototype of the application is already in place. The Furhat SDK includes a virtual version allowing for remote development. The Bachelor thesis primarily revolves around integrating social cues to be leveraged by Furhat with the help of a Large Language Model. Social cues are subtle signs, verbal or non-verbal, communicated during social interactions that provide information about the others' feelings, intentions, and expectations. In the context of Furhat, examples of non-verbal social cues are mimics and gestures. Examples of verbal cues are emotional language, humor, etc. You are also welcome to contribute your own ideas.

Skills required

• Good programming skills (e.g., Java, Kotlin, Python)
• General interest in generative AI, large language models, and/or human-computer interaction
• Good time management and organizational skills
• English skills

Contact

If you are interested and want to apply for this bachelor thesis, please contact Leon Hanschmann (leon.hanschmann∂kit.edu) with a short motivation statement, your CV, and a current transcript of records. Feel free to reach out beforehand if you have any questions.