Interviewing AI: A new challenge for discourse analysis in the neural age

This study addresses similarities and differences as to how three chatbots—namely ChatGPT, Bing Chat and Google Bard—‘converse’ with humans; in so doing, I also analyze to what extent the answers yielded by chatbots may differ from the typical answering patterns featured in human-to-human interviews. To carry out the study, I rely on a corpus specifically compiled to comprise dialogical interactions between AI and humans. Data show that these three chatbots exhibit both similarities in the way their training has been improved since the debut of ChatGPT in 2022, and differences in dealing with their interlocutors. Specifically, Bard’s replies point to a chatbot which accompanies the interlocutor without imposing itself; Bing Chat favours impartiality and neutrality and avoids falling into the trap of problematic conversation; finally, ChatGPT acts like a tutor that leads the way, rather than simply help.