Will You Accept an Imperfect AI? Exploring Designs for Adjusting End-user Expectations of AI-powered Systems

Explore the impact of expectation-setting methods on user acceptance of AI systems in this 21-minute conference talk from the ACM CHI Conference on Human Factors in Computing Systems. Delve into a study using a Scheduling Assistant AI to examine how different error focus and expectation adjustment techniques affect user perceptions of accuracy and system acceptance. Learn about the researchers' findings on preparing users for AI imperfections and the resulting significant increase in acceptance. Gain insights into the challenges of incorporating AI technologies into end-user applications and the importance of managing user expectations for successful adoption.

Intro
Motivation
Research Questions
Al Application: Scheduling Assistant
Al errors
Understanding Expectations
Expectation Adjustment Strategies
Study 1: Hypotheses
Study 1: Conditions
Study 1: Procedure
Study 1: Results
Study 2: Hypotheses
Study 2: Procedure
Study 2: Task
Study 2: Conditions
Study 2: Results
Summary of Results
Conclusion