What Makes People Say Thanks to AI

Yicong Yuan,Mingyang Su,Xiu Li

DOI: https://doi.org/10.1007/978-3-031-60606-9_9

2024-01-01

Abstract:This study delves into the dynamics between user politeness and the intelligence level of conversational AI products, alongside their interaction methods. We focused on the evolving sophistication of AI, especially in large language models like GPT, and its influence on user behavior and perception. A notable finding is the significant correlation between AI intelligence and the frequency of user politeness. As AI progressively mimics human-like understanding and interaction, users tend to engage more politely, viewing these interactions as akin to communicating with a peer. We also highlight the importance of interaction modes. While users generally show more politeness in text and voice dialogues compared to simpler interfaces, the anticipated superiority of voice dialogues in eliciting politeness over text was not observed, which suggests that the interaction format may be less impactful than the AI’s perceived intelligence. Furthermore, our identify a positive link between user satisfaction and politeness, positing that politeness could act as an indirect indicator of user satisfaction with AI products. This method offers a less intrusive alternative for assessing AI effectiveness, diverging from direct metrics like task efficiency or subjective satisfaction. Our research offers new insights into users’ polite behavior towards AI products. It reveals which product characteristics prompt users to exhibit polite behavior and highlights the significance of observing user politeness for AI product design.