Kristina Malzahn
Leibniz University Hannover, Institute of Philosophy
A widely accepted premise in the philosophy of science is that scientific images possess epistemic value, derived from their capacity to represent reality. On this image-centric view, the evidential value of a visualization derives exclusively from its visual content. An unambiguous depiction of a phenomenon is considered evidence, irrespective of its origin or application in reasoning. I reject this assumption. Scientific images are not evidence in themselves; they become epistemically significant only when integrated into practices of interpretation, methodological transparency, and criticism.
It is particularly pressing to consider artificial intelligence (AI) in scientific imaging, as in astronomy and pathology, where convolutional neural networks generate photorealistic outputs emulating traditional methods. Such visual plausibility does not guarantee epistemic warrant: an image may resemble evidence without preserving the causal or model-based relation to its object. The central thesis is that the epistemic force of such images is derivative, not intrinsic. They count as evidence only given a documented production process, assumptions open to scrutiny, and a justified role in scientific argumentation. The shift from indexical to AI-generated images makes this condition unavoidable: when the generative process is opaque, epistemic legitimacy can no longer be read off the image itself.
This implies that scientific seeing is not merely passive acceptance of facts, but a socially organized form of reasoning. Objectivity emerges from criticism, transparency, and shared standards — not from the visual surface of the image.
Chair: tba
Time:
Location:
SOPhiA 2026 