Cultural Relevance Index: Measuring Cultural Relevance in AI-Generated Images

Wala Elsharif, Marco Agus, Mahmood Alzubaidi, James She

Abstract: This paper introduces the Cultural Relevance Index (CRI), a metric designed to evaluate and quantify the cultural relevance of AI-generated images. Leveraging the detection capabilities of GPT technology and our proposed mathematical formulas aligned with human perception, CRI assesses the extent to which the content of an image is relevant to a particular culture such as Arabic, a capability not commonly found in many AI technologies. Through rigorous validation, including comparison with human judgments and a mathematical baseline, with focus on Arabic culture, CRI demonstrates a compelling similarity (around 95%) with human judgement, outperforming a quantified baseline metric (over 28%). CRI also succeeds to matches human judgment in comparing cultural images 100% of the time.