Abstract
Background: Transparency initiatives in AI systems aim to encourage data-sharing, yet sharing behaviors frequently diverge from stated intentions. This intention–behavior gap reflects dual-process dynamics: intuitive (System 1) responses guide actual sharing behavior, while deliberative (System 2) reasoning governs willingness to share. Whether transparency influences deliberative preferences without affecting immediate behavior – and how trust moderates these effects – remains unclear. This study investigates how transparency, trust, and the processing entity type (human vs. AI) differentially influence deliberative versus immediate sharing decisions, addressing a gap in understanding dual-process dynamics in AI contexts.
Method: To isolate these effects, we conducted a pre-registered online experiment (N=240) where participants interacted with a fictional sleep-optimization app. They were randomly assigned to scenarios where data was processed by either a human expert, a transparent white-box AI, or an opaque black-box AI. This design allowed testing the impact of entity type and transparency on actual data-sharing and willingness to share, while measuring the moderating roles of trust and privacy concerns.
Results: Counter to common assumptions, AI transparency alone did not significantly increase data-sharing. Its positive effect on willingness to share was contingent on pre-existing user trust in AI, particularly for white-box systems. This suggests trust enables transparency’s benefits. Moreover, actual sharing often contradicted willingness to share (the privacy paradox), with consistently high sharing rates across all conditions indicating that immediate decisions were largely driven by intuitive System 1 processing rather than deliberative evaluation.
Conclusion: This research challenges the direct benefits attributed to AI transparency in promoting data-sharing, revealing its effectiveness is amplified by, and dependent upon, user trust. It extends privacy and dual-process theories by showing intuitive System 1 processing can dominate AI data-sharing contexts, overriding stated concerns. Practically, fostering trust in AI may be a more vital prerequisite for data-sharing than implementing transparent designs.
Recommended Citation
Rosenberger, Julian; Kuhlemann, Sophie; Tiefenbeck, Verena; Kraus, Mathias; and Zschech, Patrick, "Understanding Data-Sharing with AI Systems: The Roles of Transparency, Trust, and the Processing Entity" (2026). PAJAIS Preprints (Forthcoming). 59.
https://aisel.aisnet.org/pajais_preprints/59