A Practical Tool for Responsible AI
Framework Design | User Testing | Olivia Gambelin, author & thought leader | 5 weeks
From
Book’s worth of heady information
To
Actionable framework on one tidy page
Background
Our client, Olivia Gambelin, a responsible AI thought-leader, was weeks away from a final deadline for her book manuscript. She was missing a big piece: a graphic framework to summarize her innovative model.
Challenge
How might we help business leaders answer the two biggest blockers to implementing Responsible AI:
Where do I start?
What am I missing?
Result
Designed and tested a one-page canvas published in Responsible AI in June 2024 and available as a standalone download at thevaluescanvas.com.
Team
Kirsten Collins, Strategist
Michelle Zamora, Graphic Designer
Armando Somoza, Facilitator
yves louise, Content Designer
Skills
framework design
focus group facilitation
insight synthesis
The Challenge
Olivia came to us with a full-length manuscript for her forthcoming book, Responsible AI: a step-by-step process for creating and implementing an ethical AI strategy.
With just a five weeks until her final manuscript was due, she was struggling to represent her framework visually.
Her graphics effectively represented the sections of the book, but she wanted an actionable tool for readers to implement Responsible AI in their work.
before: author’s initial graphics
Exploration
After a discovery call with Olivia, we analyzed the transcript and began experimenting with different information hierarchies and visual metaphors.
reference images of pillars, building blocks, canvas
changing the pyramid hierarchy
system framework exploration
a more detailed pyramid
funnel framework exploration
A Direction Clicks Into Place
Because Olivia desired an actionable tool, we wondered if leaving space for readers to fill in the blanks would work. Once we started exploring a canvas model, things clicked into place.
canvas framework v1
canvas framework v2
Test with Users
Despite the tight timeline, I advocated for a round of usability testing through two methods:
Survey circulated to business leaders and AI ethicists
Live scenario simulation with a focus group
simulation hosted over Zoom
users tested the canvas in a collaborative Miro board
Insights & Actions
Our research insights led to changes in the final version:
Adding space to list existing ethical values to connect the canvas to prior work
Adding a MadLib-style sentence to summarize each section
Tweaks to hierarchy and examples for greater clarity
Results
Branded “The Values Canvas,” our visual framework was published in Responsible AI in June, 2024, is available as a standalone tool at thevaluescanvas.com, and in use at companies worldwide who are committed to responsible, values-aligned strategies for artificial intelligence.
