Case Studies

The case studies share real world stories from teams who have designed AI-driven products using human-centered AI best practices.

Guidance in action

A collection of real world examples of human-centered AI design, done with or in the spirit of the Guidebook.

We add new case studies on a rolling basis.

Google Photos

The Google Photos team on balancing user control and automation in AI-powered photo memories.

Relevant patterns

Determine if AI adds value Let users give feedback Let users supervise automation Give control back to the user when automation fails

BenchSci

The BenchSci team on designing a user-centered, AI-powered biomedical research tool.

Relevant patterns

Determine if AI adds value Be accountable for errors Anchor on familiarity Let users give feedback Get input from domain experts as you build your dataset

Google Flights

The Google Flights team on finding a user-friendly way to communicate AI-powered price predictions.

Relevant patterns

Set the right expectations Explain the benefit, not the technology Determine how to show model confidence, if at all Explain for understanding, not completeness Go beyond in-the-moment explanations Give control back to the user when automation fails

Read Along

Google’s Read Along team on designing a parent and kid-friendly product to grow child literacy.

Relevant patterns

Set the right expectations Explain the benefit, not the technology Make precision and recall tradeoffs carefully Be transparent about privacy and data settings Make it safe to explore

Tune

Google’s Jigsaw team on designing a user-controlled, AI-enabled comments moderation tool.

Relevant patterns

Be accountable for errors Anchor on familiarity Let users give feedback Let users supervise automation Give control back to the user when automation fails