Projection-Induced Determinism: An Epistemic Bias in Low-Dimensional Representations of High-Dimensional Systems

Read full text Summarise with VIRAT
Authors: Bernd von Mallinckrodt
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Dimensionality reduction does not only simplify systems — it can change how they appear to behave. This paper formalizes Projection-Induced Determinism (PID) as a bias mechanism explaining how apparent predictability can emerge from projection itself. This preprint introduces Projection-Induced Determinism (PID) as an epistemic bias arising from dimensionality reduction in complex systems modeling. Across disciplines such as neuroscience, ecology, economics, and machine learning, high-dimensional systems are routinely represented in low-dimensional spaces to enable analysis, prediction, and interpretation. These representations are widely assumed to preserve the essential dynamical structure of the underlying system. This paper examines that assumption and argues that the projection from a high-dimensional state space to a reduced observable space is not neutral. Instead, it systematically alters the apparent dynamical properties of the system. PID is defined as the mechanism by which the elimination of orthogonal degrees of freedom reduces measurable uncertainty, contracts the effective state space, and produces an apparent increase in predictability that is not intrinsic to the system but arises from the observational mapping. The formal framework introduces a projection operator M: \mathbb{R}^n \rightarrow \mathbb{R}^m with n \gg m, and defines PID in terms of a reduction in conditional entropy in the projected representation relative to the full system. Three mechanisms are identified: (i) elimination of orthogonal components, (ii) geometric contraction of the state space, and (iii) an apparent increase in predictability. The concept is positioned relative to existing literature on coarse-graining, dimensionality reduction, non-Markovian projection effects (Mori–Zwanzig formalism), model misspecification, early warning signals of critical transitions, and information theory. Rather than replacing these approaches, PID provides an interpretive lens that clarifies how projection can influence the apparent causal structure inferred from reduced models. A conceptual illustration using neurobiological measurement highlights how low-dimensional observables may appear stable and deterministic while remaining coupled to unobserved dynamics. The implications extend to model interpretation, causal inference, and the evaluation of predictive signals in reduced representations. This work does not propose a new modeling framework but formalizes a general bias mechanism that constrains how results from low-dimensional models should be interpreted. Primary Keywords Projection-Induced Determinism Dimensionality Reduction Epistemic Bias Complex Systems High-Dimensional Systems Secondary Keywords Coarse-Graining Manifold Learning Non-Markovian Dynamics Model Misspecification Conditional Entropy Cross-Disciplinary Keywords Systems Neuroscience Early Warning Signals Critical Transitions Information Theory Dynamical Systems Causal Inference Machine Learning

Fields of Study & Subjects

Complex Systems and Dynamics Model Reduction and Neural Networks Ecosystem dynamics and resilience

Publication details

Journal / source
Zenodo (CERN European Organization for Nuclear Research)
Publication date
28 March 2026
Language
EN
Document type
Preprint