The dominant narrative surrounding AI describes its impact across scientific disciplines as a significant shift—one that extends beyond mere tooling to influence the very fabric of research practices and epistemic cultures (Falk 2019; Fösel et al. 2019; Gethmann et al. 2022). This narrative is rooted in new and expanded types of prediction. Such capabilities require scientific disciplines to accommodate two interrelated challenges: uncertainty and simulated data. Both of these factors will be investigated at this conference across various methodological traditions.The key question is how the impact of new AI-based research practices should be assessed.
We believe it is essential to conceptualize AI neither as a mere neutral tool nor as a magical oracle. Instead, we aim to interrogate and assess its transformative power against the background of current methodologies in various scientific disciplines. Our extensive media ethnographies of AI-based research suggest that these implementations rarely remain uncontested. How then do validation, application, and critique in specific contexts of forecasting and future orientation play out?
To accomplish this task, we do not only suggest a focus on the predictive practices which enable scientists to leverage AI to understand and increasingly anticipate complex research problems. Adopting a more critical stance, we will also interrogate the hypothesis that the increasing reliance on simulated data fosters a progressive decoupling from empirical grounding. This
perspective directly addresses the concern that AI-driven research may be developing selfreferential epistemic loops, where models are trained and validated against synthetic dataverses that are themselves simulated, rather than against direct measurement. This raises urgent questions about a potential ‘profound loss of indexicality’ and the introduction of new, deeply embedded biases originating from a simulation's core assumptions. We will therefore probe whether this shift does not merely augment traditional methods but actively redefines the status of evidence, potentially promoting a ‘blur’ or ‘illusion’ in relation to scientific truth claims. The challenge of uncertainty is thus transformed: It becomes less a question of managing known imprecision and more a fundamental doubt about the veracity of the simulated worlds from which predictions increasingly originate.