Gain insights from the People + AI Research team on building trustworthy, user-centered AI products in this 55-minute workshop. Explore opportunities in the AI development process for improving and calibrating user trust through a series of exercises. Learn about a broader toolkit of resources for further exploration, including the People + AI Guidebook and techniques for classifying text with BERT. Discover best practices for early data handling, maintaining datasets, creating realistic data, and designing user experiences that set appropriate expectations. Understand how to explain AI benefits and recommendations, handle model confidence and errors, and incorporate user feedback. Walk away with practical knowledge to develop AI products that balance innovation with user trust and ethical considerations.
Overview
Syllabus
Introduction
What is the Pear Guide Book
Chapter Updates
Case Studies
Workshop Resources
User Need
Patterns
Early good data practices
Embrace noisy data
Maintain the data set
Create realistic data
Recommendations course
Recommendation Systems course
User Experience
Design Patterns
Setting the right expectations
Explaining the benefits
Explaining recommendations
Recap
Model Confidence
Errors
Examples of errors
Let users give feedback
What action will be taken
Example
Summary
Taught by
TensorFlow
