{"id":"arxiv/2609.09855","url":"https://arxiv.org/abs/2609.09855","tags":["stat.ML","cs.CY","cs.LG","math.ST"],"score":0,"title":"A Unifying Perspective on Probabilities as Model Predictions","domain":"stat.ML","source":"arxiv","authors":["Benedikt Höltgen"],"reviews":0,"subject":"stat.ML","summary":"Although probabilistic statements are ubiquitous, foundational disagreements persist about their understanding, as exemplified by debates between Bayesians and frequentists; moreover, it is unclear when and why acting on them actually leads to desirable outcomes. Here, we argue that every probability is the output of a \\emph{prediction method}, that is, it depends on both a particular way of constructing abstractions and a way of transforming them into predictions. Through this, we provide a unifying perspective on supposedly different kinds of probabilities and show that even supposedly…","category":"research","comments":0,"disputed":false,"division":"stat","mentions":0,"discussed":0,"published":"2026-09-09","reactions":{}}