I Made the Choice, Did I? Exploring the Influence of Goal-Directed AI on Human Decision-Making

Abstract

As AI become integrated into decision-making, some concerns emerge: Can AI shape our decisions without us noticing? To examine this, we adopt forcing techniques from performance magic, which guide apparently open choices toward predetermined outcomes. This provides a lens for examining how goal-directed AI can steer decisions while preserving users’ sense of agency. We conducted an exploratory study with 14 magicians to examine forcing techniques and derive design principles for AI Magician, a research probe for goal-driven AI influence. We then performed a mixed-methods study (n = 30) comparing AI Magician with conventional AI recommendations in a design task. Our results showed that participants often incorporated predefined target attributes without identifying the specific target during the process. Despite this influence, they reported comparable agency and trust,suggesting that subtle AI influence can coexist with perceived autonomy. We discuss ethical risks and implications for transparency, agency, and safeguards in AI-mediated decision-making.