What can AI-generated 3D objects reveal about my self? Exploring the potential of AI-generated self-images for critical reflection on content consumption shaped by Netflix algorithms

Abstract

Algorithms shape content consumption yet limit self-awareness by obscuring how they model users’ identities. To address this, we explored whether generative AI–based 3D object visualizations, created from individuals’ Netflix viewing histories, could support reflective engagement with algorithmically inferred preferences. Unlike traditional data visualizations, our approach externalizes content consumption as metaphorical self-images, enabling users to interpret their algorithmic identities through visual cues. Through semi-structured interviews with ten participants, we found that AI-generated self-images revealed unconscious dispositions, evoked emotional responses, and encouraged rethinking of personal tastes. However, issues of contextual misalignment and interpretive ambiguity emerged when emotional or situational intent was absent from the input data. Our findings suggest that such visualizations can foster self-awareness by transforming behavioral data into expressive, interpretable forms. We outline design considerations to make AI-generated self-images more interpretable, engaging, and personally meaningful to better support algorithmic self-reflection.