From Quantity to Uniformity: A Conceptual Framework Linking Recursive Model Collapse and Guénon’s Critique of Quantification

This article argues that recursive synthetic-data training should be understood not merely as a problem of performance degradation, but as a potential narrowing of distributional breadth, in which rare and locally distinctive information may disappear before aggregate benchmarks reveal any decline. René Guénon’s critique of the displacement of quality by quantity is used here as a diagnostic vocabulary for this narrowing, not as a causal theory of why it occurs. A structured synthesis of secondary literature is combined with a proposed recurrence relation for mixtures of human and synthetic data and a multi-indicator monitoring framework comprising tail coverage, semantic diversity, distributional closeness, and held-out performance. The reviewed evidence is consistent with three provisional conclusions: recursive replacement of human data can reduce coverage of low-probability events; tail-sensitive degradation may precede detectable changes in aggregate benchmarks; and preserving human data is not sufficient unless the synthetic component remains within a quality-dependent range. These conclusions are not presented as new empirical findings. They constitute theoretically motivated hypotheses and a protocol for future primary research. The article’s primary contribution is conceptual and diagnostic rather than empirical or predictive: it distinguishes performance degradation from distributional narrowing, and proposes Guénon’s vocabulary as a conceptual frame for interpreting the latter, alongside a family of quantitative indicators for detecting it before the former becomes visible.