Technology paternalism and the reactance deficit — Are our natural protective mechanisms failing us?

With the rapid advancement of artificial intelligence (AI), the potential for paternalistic system behavior increases. While restrictions on autonomy are known to undermine psychological well-being, AI systems introduce a particular challenge: their opacity may prevent users from recognizing when and how their autonomy is constrained. As a result, autonomy restrictions may go unnoticed, potentially weakening natural protective mechanisms such as psychological reactance. To investigate the relationship between AI paternalism, autonomy perceptions, and reactance, we conducted a mixed-design online experiment ( N = 137) in an e-commerce context, manipulating five levels of AI influence (control, nudging, reversible restriction, non-reversible restriction, covert restriction). Autonomy perceptions and reactance were assessed before and after a belief-updating intervention disclosing the system’s paternalistic mechanisms. Results show that covert restrictions initially elicited comparatively weak reactance, whereas salient nudges triggered elevated reactance despite preserving choice. Belief updating partially recalibrated autonomy perceptions, particularly when prior restrictions had remained opaque. Mediation analyses support a dual-pathway model of technology reactance: changes in perceived personal autonomy constitute the primary mechanism underlying reactance, while increased attribution of system autonomy provides an additional pathway. These findings suggest that reactance in AI contexts depends less on restriction per se than on how autonomy threats are cognitively represented. We propose a dual-pathway framework linking paternalism perception, belief updating, and reactance to guide future research on autonomy-preserving AI design.