Overview

In 2019, Skype users were surprised to learn that Microsoft vendors were listening to voice clips from calls in order to train speech transcription models. As a company committed to building experiences that users trust, we were faced with a challenge: When we ask users for their consent to collect and use speech data, how can we better explain what privacy protections are in place, as well as the benefits of opting in? To address this challenge, Ethics & Society partnered with product teams to redesign when and how privacy choices show up for end users, with the goal of creating speech consent experiences that users understand and trust. During this session, we’ll walk you through how we approached our redesigns, and share actionable guidance for how you and your product teams can create consent experiences that respect users while still helping you get the high-quality training data needed to build high-quality ML models.