THE INVARIANT FRONTIER Deterministic Chaos, Ecological Rationality, and the Structural Limits of Scalable Artificial Intelligence

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

The prevailing narrative in artificial intelligence holds that scale — more data, more parameters, more compute — asymptotically extends the reach of predictive and optimizing systems into domains previously considered intractable. This paper argues that this narrative rests on a category error: it conflates epistemic uncertainty, which additional information can reduce, with dynamical indeterminacy, a structural property of nonlinear systems that no amount of data can eliminate. Drawing on the mathematics of deterministic chaos and a formal decision-space model distinguishing algorithmic rationality, heuristic self-organization, and randomness, we show that increasing data density sharpens the boundary between these regions without ever displacing it. A quantitative argument based on Lyapunov exponents demonstrates that improved precision buys, at best, a constant additive prediction horizon rather than expanded territory. The paper concludes with implications for organizational governance and compliance design, arguing that treating AI-assisted forecasts as if they could cross this invariant frontier constitutes a specific and under-recognized category of decision risk.