Join SoapBox + CA at SXSW EDU 2024: Learn more

Equity in
voice AI

Creating equitable voice experiences for kids

For SoapBox, equity means ensuring that every child’s voice is understood accurately regardless of individual and unique characteristics like accent, dialect, learning ability, race, socioeconomic background, or ethnicity. 

When Dr. Patricia Scanlon started SoapBox in 2013, she imagined building voice experiences for kids in a New York City classroom. That vision, based on diversity and the high stakes of an educational setting, has driven every technical, process, and design decision SoapBox has made in the last 10 years, and continues to drive our mission today.

Download Page Image 3

Equity in Voice AI Checklist

Education companies see voice AI as a powerful new feature of their literacy, language and math tools, but few know where to start when assessing voice AI for accuracy and bias. 

Enter our new checklist of questions to ask your voice provider of choice before integrating their AI into your tools for prek-12 classrooms.

Certified for mitigating bias

In October 2022, SoapBox became the first AI company to be awarded the Prioritizing Racial Equity in AI Design certification from Digital Promise and The EdTech Equity Project.

This certification recognizes our steadfast commitment to mitigate racial bias and understand all kids’ voices accurately and equally. Read our blog to learn more about this certification and our approach to building equity-driven voice AI technology for kids.

The Edtech Equity Project Logo 2

How we mitigate bias

Building our equitable voice AI technology requires training and testing our models on diverse voice data of all accents and dialects.

In this video, CTO Dr. Amelia Kelly describes one of our approaches to testing our technology for bias. It’s called “Describe the Picture.”

We’re thrilled to see SoapBox Labs deliver best-in-class voice-enabled technology that sets a high standard for the edtech industry by investing in the critical work required to ensure their AI works well for learners of color.

Nidhi Hebbar
Co-Founder, EdTech Equity Project

Frequently asked questions

How does SoapBox’s voice engine cater to different accents and dialects?

We ensure that our voice engine understands all kids’ voices equally by training our acoustic models on kids’ accents and dialects from all 50 states and from 193 countries. The SoapBox voice engine can also easily accommodate unique or custom pronunciations for any population and has been independently validated by researchers, clients, and Digital Promise to mitigate bias across race, accent, and socioeconomic background.

Describe SoapBox’s approach to developing equitable AI.

The work involved in delivering equity with our voice engine starts with how we structure our processes around modeling, building, and testing. These structures need to be proven, comprehensive, scalable, and include all categories of data for a broad range of ages, genders, ethnicities, races, use cases, environments, and much more.

SoapBox proactively identifies and mitigates bias in our speech technology by:

  1. Ensuring diversity in the data we use to train and test our system.
  2. Continuously testing, tuning, and improving our AI algorithms.

Learn more about our approach to delivering equitable voice AI.

How is SoapBox’s voice technology designed to drive equitable preK-12 learning experiences?

SoapBox has been building proprietary voice AI for kids in the classroom since 2013, and we’ve been committed to delivering voice experiences that understand every child’s voice equally from the very beginning.

It’s critical that the voice technology kids use in the classroom is equitable. Through our education partners, we’ve already delivered over 80 million learning experiences for kids across the world using our voice technology.

And the work of delivering equity is never finished. We work closely with our partners to continue to test our engine in their products. We also offer teachers the ability to listen back to their students’ voice recordings and to manually edit the voice data that populates their dashboards, so that teachers and their expertise remains at the center of the learning experience.