Kurt Weichselberger's Contribution to Imprecise Probabilities

Thomas Augustin, Rudolf Seising

Introduction: Challenging the Dogma of Precision

Classical probability theory, established through Kolmogorov’s axioms, has long operated under what Kurt Weichselberger termed the "dogma of precision." This framework mandates that the probability of any event be expressed as a single, precise numerical value. However, in complex fields such as artificial intelligence and statistical inference, information is frequently too sparse or uncertain to justify such exactitude.

Kurt Weichselberger, a pioneer in the theory of imprecise probabilities.

Throughout his career, Weichselberger sought to break this historical habit of thinking. He recognized that to maintain mathematical rigor in the face of uncertainty, the concept of probability needed to be expanded from a single point to a set of numbers—an interval. This paper explores Weichselberger’s lifelong quest to build a sound axiomatic foundation for these "imprecise" probabilities, providing a framework capable of handling uncertain knowledge without sacrificing the strengths of classical statistics.

The Axiomatic Foundation of Interval Probability

Weichselberger’s primary contribution was the rigorous axiomatization of interval-valued assignments, generalizing Kolmogorov's work. He introduced two critical concepts to distinguish the levels of consistency within these assignments:

  • R-probability (Reasonable): Assignments that avoid "sure loss" and maintain a basic level of logical consistency.
  • F-probability (Feasible): Assignments that are fully coherent within the broader set of possible precise probabilities.

Unlike many of his contemporaries, such as Peter Walley—who grounded imprecise probability in subjective betting and gambles—Weichselberger focused strictly on the probability of events. He maintained a commitment to $\sigma$-additivity and ensured his axioms remained independent of specific interpretations, whether frequentist or subjective. To achieve this, he utilized linear programming and duality results not merely as calculation tools, but as core mathematical components for his proofs.

Epistemic Symmetry and the Critique of AI Heuristics

In the 1980s, as expert diagnostic systems like MYCIN gained popularity, researchers often relied on ad-hoc heuristics like "certainty factors" that lacked mathematical backing. Weichselberger was a vocal critic of these measures, including the Dempster-Shafer combination rule, arguing that they compromised statistical rigor. He demonstrated that Probability Intervals (PRIs) could offer the necessary flexibility for AI without undermining the system's logical integrity.

One of his most profound insights involved the distinction between epistemic symmetry and physical symmetry. In his framework, epistemic symmetry (the absence of knowledge regarding asymmetry) is modeled by a continuum of uniform probabilities. In contrast, physical symmetry (the active knowledge that a system, like a fair coin, is symmetric) justifies the use of traditional, precise uniform probabilities. This distinction allowed for a more nuanced modeling of hierarchical Bayesian inference without the need for arbitrary prior assumptions.

Legacy and the HiStaLMU Project

Weichselberger’s research program reached major milestones with the publication of A Methodology for Uncertainty in Knowledge-Based Systems (1990) and his magnum opus, Elementare Grundbegriffe... (2001). These works provided the Kolmogorov-style foundation required to move interval probability into the mainstream of statistical theory, including the development of Neyman-Pearson testing under general interval conditions.

Weichselberger with colleagues, illustrating the collaborative history of his research program.

Despite these achievements, Weichselberger’s vision remained partly unfinished at the time of his death in 2016. Several volumes of his work and early manuscripts on logical probability remain lost or unpublished. Today, the HiStaLMU (History of Statistics at LMU Munich) project is dedicated to preserving his estate and conducting bibliometric analyses of his influence. The ongoing archival work suggests that a detailed rework of his unfinished manuscripts may yet trigger a paradigmatic shift in how we understand probability and statistics in the modern era.

Summary updated

Abstract

Kurt Weichselberger, one of the influential senior members of the imprecise probability community, passed away on February 7, 2016. Almost throughout his entire academic life the major focus of his research interest has been on the foundations of probability and statistics. The present article is a first attempt to trace back chronologically the development of Weichselberger’s work on interval probability and his symmetric theory of logical probability based on it. We also try to work out the intellectual background of his different projects together with some close links between them.
BibTeX
@inproceedings{augustin2017kurt,
  author = {Thomas Augustin and Rudolf Seising},
  title = {{Kurt Weichselberger's Contribution to Imprecise Probabilities}},
  booktitle = {ISIPTA},
  year = {2017},
}
Figure