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

This paper introduces AI replicas as an innovative alternative to traditional anonymization methods in image-based qualitative research. It highlights the ethical and practical dilemmas posed by conventional anonymization methods, such as the distortion or loss of emotional and contextual information, and proposes AI replicas to preserve the integrity of visual data while ensuring participant anonymity. The paper explains the use of generative artificial intelligence (AI), particularly Stable Diffusion, to create AI replicas for anonymization and fictionalization purposes. Additionally, it addresses potential biases in generative AI and offers strategies for mitigation through careful prompt engineering and participatory approaches. This approach aims to enhance ethical practices in visual research by maintaining participant anonymity without compromising the data's qualitative richness and interpretative validity.