From Archives to Algorithms: Distance, Evidence, and Inference

This article reinterprets Carlo Ginzburg’s indiciary paradigm as a general theory of knowledge production and connects it to contemporary debates over generative artificial intelligence. In line with Ginzburg, I posit that we cannot directly access unmediated social life. But rather than treat distance as an obstacle to knowledge, temporal, epistemic, and perspectival forms of distance are its enabling conditions. We can make sense of these by weighing positive and negative analogy transfers (Mary Hesse) between radically different form of knowledge traces. Consider for example archival records, or the outputs of in silico research. Both domains require reasoning from traces that stand in for absent realities. Yet, synthetic outputs derive their authority from optimization and their plausibility is operational, rather than referential. A case study of Nazi racial science clarifies what is at stake when AI systems are treated as stand-ins for social actors, and shows how perspective can be abstracted from subjecthood and redeployed instrumentally: the extraction of epistemic resources without reciprocity, and the obscuring of production processes. I introduce the concept of in silico perspectivism to name a reflexive methodological stance adequate to this moment.