ListenUp: A Gender Inclusive Speech Technology Training Tool

Background

Identical speech can be evaluated differently depending on the perceived gender of the speaker. Women’s voices are often judged by a double standard, perceived as either too insecure or too confident. Because speech perception is rapid and largely automatic, these judgments can be difficult to notice and change, even for listeners who want to be fair.

To mitigate this bias, women are often advised to lower their voices or adapt their speaking patterns in professional settings. However, such strategies rarely work consistently and may even backfire by undermining perceived authenticity. They also place additional communicative labour on women.

Project goals / research focus

ListenUp shifts attention from the speaker to the listener. It will develop and test a lightweight, browser-based listening intervention that helps users become aware of voice-based gender bias and practice listening in a more reflective and inclusive way.

Adapting state-of-the-art neural speech synthesis and voice conversion technologies, the project will create controlled speech stimuli that systematically vary voice identity and relevant voice cues (e.g., intonation patterns or voice quality).

The project will produce a tested ListenUp demonstrator, a pilot evaluation focused on change and impact, and a publication-ready study to support future development and broader implementation.

Advancing gender equality

The long-term aim is to reduce the everyday penalties women face for sounding ”too uncertain” or ”too assertive,” helping their contributions be taken seriously on equal terms without requiring them to adapt their voices to fit narrow expectations.

Researcher

The project is led by Eva Székely, PhD in Technology and researcher at the Department of Speech, Music, and Hearing (TMH), EECS School, KTH.