Advances in computer vision, machine learning, and deep learning have informed new methodologies in the humanities for over a decade, but the uptake of AI models for visual analysis in art history has been piecemeal, and recent discourses have been dominated by computer scientists. This presentation explores some of the tensions, challenges, and possibilities that are emerging at this disciplinary nexus. Specifically, the presentation considers how to create and implement computational tools that preserve the insights of historically and theoretically informed art history. How might researchers mediate between computational data and humanistic reasoning? What kinds of conceptual framing are required to ensure the interpretative value of quantitative outputs? Various examples of AI-assisted interpretation are explored and considered in relation to human, curiosity-led observation. Can art history can preserve the methodological plurality of the discipline while also facilitating innovations in AI?
