Impact of Users' Beliefs in Text-Based Linguistic Interaction

Abstract

Linguistic interaction between humans and machines is one of the most challenging fields in the development of next-generation User Interfaces. In this work, we investigate the role of beliefs about the interlocutor in human-computer linguistic interaction. First, we introduced an experimental setup that makes use of filtered and post-processed web content to generate a realistic, generic linguistic interaction. Then, we collected dialogues from two different sets $\alpha $ and $\beta $ , corresponding to users being unaware or aware of the artificial nature of the interlocutor, respectively. The results thus obtained, analyzed using a standard t-test procedure ( $N=30$ ), demonstrate a statistically significant difference between the two sets in some of the linguistic features selected, i.e., sentence length and the number of adjectives, providing further insights to expand some of the evidence previously found in the literature.