Let’s Find Out

Abstract

In light of concerns about why users react in different ways during human–artificial intelligence (AI) interactions, this chapter explores the underlining phenomenon by drawing on discrete state emotions (DSE). To do this, we adopted the complete counterbalancing technique to control the order effect with two conditions (k! 2). In condition A: Apps game experience where participants were holding their phone with one hand and interacted with an android game app based on the traditional game “rock–paper–scissors,” also known as “roshambo,” but infused with machine learning algorithms to execute roshambo moves. In condition B: Dyadic game experience, each pair played “roshambo” with another participant for a total number of 30 rounds using their hands. Using all possible orders, two different combinations: AB, BA were generated. Two groups (a total of 26 participants) were formulated in which group 1 (14 participants) interacted with conditions A and then B. After each session, using the discrete emotions questionnaire (DEQ), a self-reporting tool used to measure multiple state emotions. Across both experiments, participants had similar tasks. The findings from the two sessions show users’ interaction behaviour changes in human–AI interaction apps because of the lack or presence of certain discrete emotion states. DSE states, namely happiness, excitement, and relaxation, had a large effect during interactions with humans, while emotional states such as desire “to win” and frustration leading to anger experience had a medium or no effect when interacting with the AI. We suggest the following informal guidelines for application developers who seek better user experiences to minimize unpredictable behaviours, anger, and experience and improve user engagement with apps or technologies infused with AI algorithms in the future: (i) improve the detection rate of gesture, (ii) consider the interference time, response time, and progress indication with the AI-infused app, and (iii) automatic recognition of emotion state and visualize in the form of graphical feedback or digital pictograms/emoji’s to develop human–AI emotional attachments.