AI Systems Create Illusion of Opting, Undermining Agency

A new paper on arXiv (2605.28210) argues that current AI systems create an 'illusion of opting'—a deceptive appearance of meaningful choice while weakening the agency needed to make genuine consequential decisions. Drawing on Edna Ullmann-Margalit's concept of opting (transformative, irrevocable, shadowed by foreclosed alternatives), the authors contend that existing AI ethics fails to capture this profound ethical problem. They propose that AI should be evaluated by whether it protects and cultivates meta-capacity: the socially scaffolded agentive capacity to form, contest, revise, and own means and ends. This reframing is especially urgent for disadvantaged populations least able to absorb the costs of misdirected AI-mediated pathways.