Research

Testers include people who are blind or have low vision, Deaf and hard-of-hearing people, neurodivergent people, older people and people who speak English as an additional language, in Australia, the United Kingdom and Ireland. Each session holds several streams at once: the screen, the voice and the task the person was trying to finish. Across many people and many products, that becomes something rare: evidence of how people the technology industry seldom studies actually use what it builds.

We love the technical problems this raises, and the chance to use technology to amplify voices that are too often overlooked. We also know the dataset exists because people chose to show us how they use technology. We treat that as a privilege and a responsibility. We use it in aggregate, within the consent each tester gave, and we want it to do some good.

They address problems we care about: making sure findings can be traced to what a real person actually did and said, treating assistive-technology users fairly in how evidence is read, protecting people’s privacy at scale, checking that accessibility holds up in people’s real settings, and verifying generative assets before anyone relies on them.

Where a method does more good shared than owned, we publish it openly instead, so anyone can use it and it stands on the public record as prior art.