When the Black community speaks, we don’t just talk—we tell stories, code meanings, flip syntax, and flex generations of cultural brilliance. But until now, artificial intelligence hasn’t really listened.
Howard University and Google Research just released a groundbreaking dataset: over 600 hours of African American English (AAE), recorded across 32 states, capturing the nuance, rhythm, and beauty of Black speech. It’s called Project Elevate Black Voices, and it aims to finally train speech recognition models to hear us as we are—without making us switch up, tone down, or “fix” our voice.
“African American English has been at the forefront of U.S. culture since almost the beginning of the country,” said Dr. Gloria Washington, co-lead researcher at Howard. “Voice assistant technology should understand different dialects… It’s about time.”
For too long, AI systems—from Siri to Alexa to customer service bots—have misheard, mistranslated, or outright ignored Black voices. Why? Because those systems were trained on datasets that didn’t include us. That’s not just a tech oversight. That’s digital erasure.
Community at the Core
This wasn’t a sterile lab experiment. The Howard team traveled across the country, hosting local events with Black panelists to lead open conversations about AI, voice tech, and data privacy. Then, they invited people to share their voices—authentically and on their terms.
Dr. Lucretia Williams, a Howard researcher and community-based project lead, made it plain: “I wanted to create a safe and trusted space for people to share their experiences and ask uncomfortable questions about tech and AI.”
The result? A data set that centers—not studies—Black people. One that’s rooted in consent, community, and cultural integrity.
What Happens Next
Google plans to use the dataset to make its voice recognition products more accurate and inclusive. But Howard will retain ownership and licensing rights, ensuring it’s used responsibly—and only by those aligned with the values of equity and empowerment.
At first, the dataset will only be available to researchers at HBCUs. That’s intentional. It ensures the data serves us before it’s shared more broadly. Access for others? That’s TBD, based on alignment with community-driven goals.
Dr. Courtney Heldreth from Google called the partnership a “tremendous and personal honor,” saying, “I believe our work here will allow more users to express themselves authentically when using smart devices.”
Why It Matters
This is about more than voice assistants. It’s about representation in the digital age. It’s about correcting course before AI becomes yet another system that misunderstands us—and profits anyway.
Because if the future is voice-first, we deserve to be heard first, too.
