The Ownership Cost of AI Goal-Setting: Empirical Evidence for a Quality-Motivation Dissociation and Differential Harm to Vulnerable Users (Extended Abstract)

Abstract

AI systems are now deployed at scale to author personal goals on users' behalf in coaching platforms and productivity tools, a practice justified by the assumption that better-structured outputs produce better outcomes. We provide a preregistered experimental test of this assumption in personal goal-setting, a domain where outputs must drive sustained behavior rather than merely be delivered. In a between-subjects experiment (N = 470), LLM-authored goals scored dramatically higher on objective SMART criteria (d = 2.26), yet participants reported lower psychological ownership (d = 1.38), commitment (d = 1.19), and perceived importance (d = 1.13). At two-week follow-up, 72.8% of self-authoring participants had acted on two or more goals, versus 46.6% in the LLM condition, a behavioral gap that goal quality cannot explain. Mediation analyses identify psychological ownership, not structural quality, as the mechanism behind every motivational and behavioral outcome. The findings reveal an equity paradox: the ownership costs are largest for users lowest in self-efficacy, the population most likely to seek AI assistance. Optimizing for output quality in identity-relevant, behavior-dependent tasks can silently degrade human Ownership and widen inequity. Authorship preservation is an ethical requirement for AI tools deployed in self-regulatory and wellbeing contexts.