Research shows that men often interrupt when communicating. Johns Hopkins engineers have discovered that this behavioural pattern applies to AI voice assistants, with men twice as likely to interrupt them. This pattern has sparked conversations on the design of these assistants, which often possess 'feminine' characteristics such as apologetic behavior and warmth, potentially reinforcing gender biases. The research underlines the need for gender-neutral design in voice tools. 

Amama Mahmood, the study lead, emphasised the impact of the ubiquity of these AI voice assistants. The way we interact with these technologies, often tinged with unconscious biases, can reshape both our relationship with technology and our social dynamics.

In a study conducted with 40 participants using a voice assistant simulation, clear stereotypes were observed in how users perceived and interacted with the AI. Male users interrupted more during errors, and responded more socially to a feminine-voiced assistant, suggesting a preference for a female voice interface.

Interestingly, a gender-neutral voice assistant that apologised for its errors resulted in fewer impolite interactions. Although perceived as less warm and more 'robotic,' Mahmood says the design of agents with neutral traits and well-considered error mitigation strategies can foster more respectful and effective interactions.

Looking ahead, the research team plans to design voice assistants that can detect bias and adjust in real-time for a more equitable interaction. They aim to include more non-binary individuals in future studies, reinforcing that voice assistant and AI design should promote equality and not reinforce harmful stereotypes.

- CyberBeat