AI-Replicas as Ethical Practice: Introducing an Alternative to Traditional Anonymization Techniques in Image-Based Research

This paper introduces the use of AI-replicas as an alternative to traditional anonymization methods in image-based qualitative research. It emphasizes the ethical and practical dilemmas posed by current anonymization methods, such as distortion or loss of emotional and contextual information in images, and proposes the use of AI-replicas to preserve the integrity and authenticity of visual data while ensuring participant anonymity. The paper outlines the technological foundations of generative artificial intelligence (AI) and the practical application of Stable Diffusion to generate AI-replicas for anonymization and fictionalization purposes. Furthermore, it discusses the potential biases present in generative AI to suggest ways to mitigate these biases through careful prompt engineering and participatory approaches. The introduced approach aims to enhance ethical practices in visual research by providing a method that ensures participant anonymity without compromising the data's qualitative richness and interpretative validity.