|2’ read||Practical UX advice for consumer-facing ML projects based on People + AI guidebook from Google User Research team|
This is based on a talk I gave for ML Fribourg meetup on UX for machine learning projects. I draw the content mainly from People + AI Guidebook, a great resource to reflect on the specificity of user experience for AI products.
As my audience was made of data scientists rather than UX specialists, I framed the tips as issues they might encounter while developping their projects. My goal was to show how a UX approach could help them find solutions.
Here is a one-line summary of each tip:
#1: I have tons of ideas: how UX can help me choose the best one to build?
Use the Triptech method (survey + focus group) designed by Google UX researchers to assess the desirability of your ideas.
#2: Do I have all the data I need?
Start from the user profiles (personas), their needs and actions and work backward to identify the data, features and labels you will need. Don't limit yourself to existing or easily available data!
#3: My users don't trust my ML-powered app, what can I do?
Help users calibrate their trust by being more transparent on data sources and confidence level.
#4: My AI virtual assistant gets abused by angry users
Human intelligence is not (yet?) the right mental model for AI: it leads to overblown expectations and frustrations. Build on existing mental models and set realistic expectations.
#5: People use my tool in all sorts of weird situations I can't even imagine
Users will always repurpose or attempt to misuse a product: plan for it and provide fail-safe solutions.
You can find more details and all the resources in the 5 UX tips for ML projects slides.