How to Compliment a Human – Designing Affective and Well-being Promoting Conversational Things

Abstract

With today’s technologies it seems easier than ever to augment everyday things with the ability to perceive their environment and to talk to users. Considering conversational user interfaces, tremendous progress has already been made in designing and evaluating task oriented conversational interfaces, such as voice assistants for ordering food, booking a flight etc. However, it is still very challenging to design smart things that can have with their users an informal conversation and emotional exchange, which requires the smart thing to master the usage of social everyday utterances, using irony and sarcasm, delivering good compliments, etc. In this paper, we focus on the experience design of compliments and the Complimenting Mirror design. The paper reports in detail on three phases of a human-centered design process including a Wizard of Oz study in the lab with 24 participants to explore and identify the effect of different compliment types on user experiences and a consequent field study with 105 users in an architecture museum with a fully functional installation of the Complimenting Mirror. In our analyses we argue why and how a “smart” mirror should compliment users and provide a theorization applicable for affective interaction design with things in more general. We focus on subjective user feedback including user concerns and prepositions of receiving compliments from a thing and on observations of real user behavior in the field i.e. transitions of bodily affective expressions comparing affective user states before, during, and after compliment delivery. Our research shows that compliment design matters significantly and using the right type of compliments in our final design in the field test, we succeed in achieving reactive expressions of positive emotions, “sincere” smiles and laughter, even from the seemingly sternest users. We conclude by providing an outlook of our contribution for the new age of large language models and prompt engineering.