Evidence-Weighted Belief States for Classification in Collider Physics

Collider physics does not usually move from raw detector signals to definite particle labels in one step: it builds claims from traces. Tracks, showers, vertices, jet shapes, missing momentum, invariant masses, and timing information are turned into particle identities, jet flavours, event processes, resonance candidates, and possible anomalies. Each step involves uncertainty. Some signals are strong. Others are partial, conflicting, or vague. A bottom jet may look partly like a charm jet. A Higgs-to-bottom-quark event may resemble ordinary quantum chromodynamics background. A heavy-baryon candidate may be an actual resonance, or just a lucky combination of tracks. An anomalous event may be new physics, rare known physics, detector noise, or insufficient modelling. Modern collider analyses often retain continuous discriminants, likelihoods, working points, systematic-uncertainty models and sample-level statistical inference rather than relying only on hard labels. They are more like evidence-weighted belief states about an event. A reconstructed object or event is not simply assigned to one category. It carries graded support for several possible interpretations, together with information about missing evidence, conflicting evidence, detector limits, and analysis choices. Quantum Epistemics (QE) offers a useful language for the knowledge-production pipeline, including the scientists’ interpretation of the results. It treats classification as fuzzy, probabilistic, and shaped by the observer-system relation, where the “observer” in collider physics means the detector, reconstruction software, simulation model, selection rule, analysis goal, and the research team. Concepts such as similarity, dissimilarity, no-match outcomes, and observer-dependent tolerance can help make ambiguity explicit rather than hiding it inside a single score. The proposed contribution is not to replace standard high-energy physics methods. Detector calibration, simulation, control regions, likelihood fits, machine learning, and systematic uncertainty studies remain the way they are. QE can help represent how detector traces become physical claims, why classifications in the human mind can be unstable, and when an object should remain unresolved rather than forced into the nearest known class. QE could provide a unified vocabulary for unresolved, missing, conflicting, and context-dependent evidence. In this sense, ambiguity and vagueness are not merely debris around a clean measurement. They are part of the evidence from which collider knowledge is made.