Accent and Gender Bias in Perceptions of Interactive Voice Systems

Abstract

Interactive Voice Systems (IVSs) are automated answering systems with pre-recorded menu options that are navigated by the user with keypresses and vocal responses. These systems are increasing in popularity with many companies because they facilitate the fielding of a large volume of callers, reduce personnel costs, improve efficiency, and increase callers’ privacy; however, do people enjoy interacting with these systems? Individual characteristics, such as accent and gender, impact human interactions through biases, and it is likely some biases are maintained in interactions with voice systems. In the present study, participants were given scenarios and instructed to interact with the IVS to complete tasks. The accent and gender of the IVS were manipulated between scenarios, and perceptions of the IVS were gathered after each interaction in terms of pleasantness, likelihood of task completion, likelihood of recommending the system to a friend, and likelihood of using the system again. It was hypothesized that the results would align with findings relating to human interactions and bias, such that participants would prefer their native accent, and the female voice would receive more negative feedback. In contrast to the hypothesis, the Mexican accent was rated more pleasant, likely to me recommended to a friend, and likely to be used again than the American accent, regardless of the participants’ native language. Overall, scenarios with positive outcomes were preferred over those with negative outcomes in all measures, and multiple interactions were found including the participant’s first language and gender and the accent and gender of the IVS.