Controllability and User Satisfaction in AI-Based Image Editing Tools

Abstract

AI-based image editing tools increasingly support automatic background removal, style transfer, and object enhancement. Despite high algorithmic accuracy, user satisfaction varies widely across application contexts. This paper examines how designers and casual users interact with intelligent editing functions during real-world tasks. Focus is placed on correction frequency, manual override behavior, and perceived controllability. Observations indicate that users often prefer slightly imperfect but adjustable outputs over fully automatic transformations. Interface transparency and reversible operations strongly influence long-term adoption.