From Data to Ficta: A Critical Reflection on Visual Analytics in the Age of Generative Models

Abstract

Over the last decades, the pervasiveness of data and the rate at which they are produced have propelled the rapid growth of Visual Analytics (VA) methods for interactive analysis and decision-making. This foundation is now shifting as computers increasingly not only process data, but also produce them through generative models. Revisiting the notion of synthetic data, we oppose data to capta (observational inscriptions) and introduce ficta: model-produced outputs that enter VA sessions as an analytic substrate with their own means of interactive production. In ficta-driven workflows, the substrate of analysis is no longer a static representation from which knowledge can be distilled; interaction becomes entangled with what is available to observe, and visualizations come to mediate not a single world but a plurality of model-admissible alternatives. We argue that this breaks VA’s classic data-driven epistemic contract and motivates a reevaluation of core assumptions behind knowledge production in the age of generativeness.