The Style of History and the History of Making: Generative AI, Painted Surfaces, and Forensic Latency

Abstract

This article examines how production histories become differently available as evidence in painted surfaces and diffusion-generated images. Drawing on Peirce’s account of the sign as a relation among sign, object, and interpretant, it develops the concept of forensic latency to describe production histories that stand in real connection to an image but can be turned into evidence only by particular interpreters, with particular means, in particular forums. In painted surfaces, such histories are lodged largely in the object and recovered mainly by examining it. In diffusion-generated images they are distributed across data, models, prompts, and platforms, and recovered, if at all, through records, re-running, or statistical inference. Generated craquelure is an icon of an index for some interpreters but itself an index of a generative system and its training images. The restoration of the Ghent Altarpiece and two disputes involving image generators show what follows for attribution, authorship, and accountability.