The Subversive AI Acceptance Scale (SAIA-8): A Scale to Measure User Acceptance of AI-Generated, Privacy-Enhancing Image Modifications

DOI

``Subversive AI'' (SAI) image filters employ adversarial machine learning techniques to protect users from facial recognition by distorting personal images in ways that confuse computer vision algorithms. While novel and of significant research interest, SAI filters are presently only evaluated in terms of attack efficacy; there is no simple way to measure user acceptability of filter outputs and, thus, researchers do not consider user acceptability when making advances. We addressed this limitation by creating and validating a scale to measure user acceptance --- the SAIA-8. In a three-step process, we apply a mixed-methods approach that closely adhered to best practices for scale creation and validation in measurement theory. Initially, to understand the factors that influence user acceptance of SAI filter outputs, we interviewed 15 participants. Interviewees disliked SAI filter outputs because of a perceived lack of usefulness and conflicts with their desired self-presentation. Using insights and statements from the interviews we generated 106 potential items for the scale. Employing an iterative process with 215 crowd-sourced participants, we arrived at the eight items that comprise the SAIA-8. Finally, through a convergent validity study with 30 participants, we found that the SAIA-8 is suitable for measuring user acceptability of privacy-enhancing image perturbations. Moreover, it can aid in prioritizing user acceptability when developing and evaluating new SAI filters.