Abstract:Commercial voice assistants are largely feminized and associated with stereotypically feminine traits such as warmth and submissiveness. As these assistants continue to be adopted for everyday uses, it is imperative to understand how the portrayed gender shapes the voice assistant's ability to mitigate errors, which are still common in voice interactions. We report a study (N=40) that examined the effects of voice gender (feminine, ambiguous, masculine), error mitigation strategies (apology, compensation) and participant's gender on people's interaction behavior and perceptions of the assistant. Our results show that AI assistants that apologized appeared warmer than those offered compensation. Moreover, male participants preferred apologetic feminine assistants over apologetic masculine ones. Furthermore, male participants interrupted AI assistants regardless of perceived gender more frequently than female participants when errors occurred. Our results suggest that the perceived gender of a voice assistant biases user behavior, especially for male users, and that an ambiguous voice has the potential to reduce biases associated with gender-specific traits.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is: when voice assistants make mistakes, how do their gender characteristics (manifested as female, male or neutral) and error - mitigation strategies (apologizing or compensating) affect users' perception of AI assistants and their interactive behaviors? Specifically, the research aims to explore the following questions:
1. **How does gender bias manifest itself when users interact with voice assistants?**
- In particular, in the context of error - prone interactions, whether users' genders will affect their views and behaviors towards voice assistants.
2. **How is the effectiveness of error - mitigation strategies modulated by the gender exhibited by voice assistants?**
- For example, between the two strategies of apologizing and compensating, which one can make users feel more satisfied or find the assistant more approachable in differently gendered voice assistants.
To answer these questions, researchers conducted an experimental study (N = 40), manipulated the gender of voice assistants (female, neutral, male) and error - mitigation strategies (apologizing, compensating), and studied how these factors and participants' genders (female and male) jointly affect people's perception and behavior towards voice assistants with errors.
### Research Background
Modern voice assistants (such as Alexa and Siri) are widely used for auxiliary tasks in various personal and professional scenarios, including information retrieval, setting reminders, sending text messages, making phone calls, etc. However, these data - driven deep - learning models are error - prone, and if not mitigated and managed, may cause users to lose trust in AI assistants. Therefore, it is crucial to study how to effectively mitigate these errors.
### Main Findings
- **Apologizing can trigger more social responses from users than compensating**: When voice assistants mitigate errors by apologizing, users show more non - verbal responses.
- **Male participants interrupt voice assistants more frequently**: Regardless of the gender of voice assistants, male participants interrupt voice assistants more frequently when errors occur, while female participants do so less often.
- **Male participants prefer feminized voice assistants that "apologize"**: Compared with masculinized voice assistants, male participants are more inclined to have more social interactions with feminized voice assistants that can apologize.
- **Ambiguously gendered voice assistants may reduce gender bias**: Using ambiguously gendered voice assistants can reduce biases related to specific gender traits.
In conclusion, this study shows that the gender characteristics of voice assistants significantly affect users' behaviors, especially more obviously among male users. Meanwhile, apologizing helps to enhance the approachability of voice assistants, and compensating can increase users' satisfaction with service recovery.