Designing AI That Delights: Lessons from Aria 2 and Spatial Audio 🎧

The best technology doesn't just work. It 𝘧𝘦𝘦𝘭𝘴 right. After years creating human-centered AI experiences, I've found they follow this hierarchy:

1️⃣ 𝗨𝘀𝗮𝗯𝗹𝗲: Does it function?

2️⃣ 𝗨𝘀𝗲𝗳𝘂𝗹: Does it solve a need?

3️⃣ 𝗩𝗮𝗹𝘂𝗮𝗯𝗹𝗲: Does it address something important?

4️⃣ 𝗗𝗲𝗹𝗶𝗴𝗵𝘁𝗳𝘂𝗹: Does it create joy?

Take Aria 2's spatial audio navigation: a breakthrough in making indoor spaces accessible. GPS works well outdoors, but navigating grocery stores, convention centers, museums, and office buildings remains a challenge, especially for blind and low-vision users. We created a solution that allows user to navigate with confidence and independence - never feel lost again. See the magic in action (video).

What made this solution successful:

𝗦𝗼𝗹𝘃𝗶𝗻𝗴 𝘁𝗵𝗲 𝗨𝗻𝘀𝗼𝗹𝘃𝗲𝗱: We discovered most navigation solutions stop at the door through extensive market research and user interviews.

𝗣𝗮𝗿𝘁𝗻𝗲𝗿𝗶𝗻𝗴 𝗳𝗼𝗿 𝗜𝗺𝗽𝗮𝗰𝘁: Collaboration with Meta's Accessibility team and Envision was crucial for authentic insights and expertise that shaped our approach from day one.

𝗗𝗲𝘀𝗶𝗴𝗻𝗶𝗻𝗴 𝗳𝗼𝗿 𝘁𝗵𝗲 𝗘𝗱𝗴𝗲𝘀: By optimizing for blind and low-vision users first, we created better experiences for everyone (the curb-cut effect in action).

𝗖𝗼-𝗖𝗿𝗲𝗮𝘁𝗶𝗻𝗴 𝘄𝗶𝘁𝗵 𝗨𝘀𝗲𝗿𝘀: We overcame logistical hurdles to test with blind users in real-world environments, calibrating design attribute real-time based on user feedback. Having a fungible prototyping platform with modular design made this process possible - kudos to our engineering team!

𝗥𝗲𝗮𝗹-𝗪𝗼𝗿𝗹𝗱 𝗧𝗲𝘀𝘁𝗶𝗻𝗴: Our biggest breakthrough is the velocity-based beacons that adapt guidance based on movement speed rather than fixed distances. It makes navigation feel natural.

The results speak volumes:

• Blind users navigated our DC office's narrow 3.5ft hallways with virtually no learning curve

• 90% of sighted users preferred our navigation prototype over existing solutions #CurbCut

• When a tester unexpectedly ran through a walking-designed space, the system adjusted dynamically thanks to input from our prior testing

To my fellow builders and designers: making AI work is table stakes. Making it intuitive and delightful is what creates truly transformative experiences.