WEB DETECTION METHODS FOR CONTEXTUAL INTERPRETATION OF IMAGE COLLECTIONS

Abstract

This paper introduces the innovative use of web detection to overcome a methodological challenge in visual media analysis within social media research and Internet studies. Traditional approaches often depend on manual coding of top-ranked images or examining individual images, lacking the integration of web-based knowledge. This study employs web detection methods utilizing image search ranking mechanisms and knowledge graphs for data retrieval and contextualization. This approach enables the identification of web entities linked to images, webpages, and image URLs that fully or partially correspond to the original image collection. Findings reveal that web detection methods offer a novel framework for examining issue mapping and cross-platform visual vernaculars, as web entities and exact visual match outputs position an image collection within the freshest and most relevant web sources. Findings are organized into key themes characterizing the (1) contextual, (2) social perceptions, (3) ephemerality and (4) technological grammar of web entities and exact visual matches. A proof of methods is provided through a case study on ChatGPT. Methodological insights are supported by a four-year digital methods study, using three unchanged image datasets to analyze AI web detection outputs over time. We applied quali-quanti approaches that facilitate a deep understanding of web detection technologies within their operational logic and socio-technical contexts. This paper contributes to new, reproducible methods for contextual image collection analysis, including techniques for studying cross-platform visual vernaculars and reconceptualizing ‘operational images’ (Parikka, 2023) through web detection methods.