The paper addresses how the basic reproduction number (R0) conflates intrinsic pathogen transmissibility with adaptive human contact behavior. Traditional interpretations of R0 often assume constant contact rates, but adaptive behavior — changes in contacts driven by perceived risk — can alter observed epidemic dynamics. The authors use directly measured mobility as an observable proxy for contact behavior to separate, to the extent possible, biological and behavioral contributions to epidemic spread.
This work has three principal aims: (i) to estimate a behavioral response function directly from mobility data; (ii) to demonstrate that the decomposition of epidemic dynamics into biological and behavioral components is not identifiable from a single epidemic trajectory; and (iii) to characterize how observationally-equivalent decompositions diverge when used for counterfactual projections.
Mobility data are treated as an empirical, observable proxy for contact rates and therefore for the behavioral component of transmission. The authors report that they directly estimate a behavioral response function from these mobility measurements rather than imposing a theoretical contact rule. The source excerpt does not report detailed methods, estimation procedures, parameter values, or the specific mobility data sources used; those methodological details were not provided in the available text.
Using mobility as a behavioral observability leverages the idea that reductions or increases in movement and visits to public places reflect changes in contact frequency and hence in effective transmission potential. Estimating how mobility responds to epidemic state is necessary to quantify the behavioral correction to any reproduction-number estimate.
A central claim is that the biology/behavior decomposition — that is, separating intrinsic transmissibility from behavioral changes in contacts — is not identified from the observed epidemic trajectory alone. The authors show that an observational-equivalence class exists: multiple combinations of biological R0 and behavioral response functions can fit the factual epidemic curve identically.
One endpoint of that equivalence class is a model with apparent constant-contact R0. Although such a model reproduces the observed trajectory, it can diverge substantially from other decompositions under counterfactual scenarios. Thus, epidemic trajectories by themselves do not uniquely reveal how much of transmission dynamics is due to pathogen biology and how much is due to adaptive behavior.
The authors name the divergence between observationally-equivalent decompositions and their counterfactual forecasts as "what R0 deletes." They characterize this divergence as state-dependent and unimodal in counterfactual severity: the magnitude of the difference varies with the assumed counterfactual epidemic state, peaking at an intermediate severity and vanishing when behavior saturates.
Behavioral saturation refers to contexts in which adaptive contacts cannot be further reduced or increased — for example, when compliance is maximized or essential contacts dominate. In such saturated states the behavioral correction to R0 effectively disappears, and the equivalence among decompositions yields similar counterfactual predictions.
Applying the framework to US jurisdictions, the authors report an empirical constraint on the behavioral correction to R0. They find a negative correlation between risk-responsiveness and behavioral non-saturation (reported r = -0.57, n = 51). In practical terms, where behavior had capacity to compensate for risk it was already near maximal (saturated), and where behavior was not saturated it showed limited responsiveness to risk signals.
Because risk-responsiveness and behavioral non-saturation are confounded in this way, the authors conclude that the correction to R0 attributable to behavior was empirically bounded and relatively modest in the jurisdictions analyzed. The excerpt does not provide jurisdiction-level results, confidence intervals, or further statistical detail; those were not reported in the source text provided here.
The study highlights a structural limitation in interpreting R0 from observed epidemic curves: without independent behavioral observables, biological transmissibility and adaptive contacts cannot be uniquely separated. Using mobility as an observable proxy allows direct estimation of behavioral response functions and makes explicit the set of observationally-equivalent decompositions.
Although the theoretical non-identifiability is real and can be characterized (including its state dependence and unimodal form), the authors report that in the empirical COVID-19 mobility data across US jurisdictions the magnitude of the behavioral correction to R0 was modest. The framework explains why this empirical modesty occurred: where behavior could offset risk it was already saturated; where it was not saturated it did not respond substantially.
Limitations apparent from the source excerpt include that methodological details, precise data sources, estimation techniques, and fuller numerical results are not reported in the provided text. Readers should also note this work is a preprint and has not undergone peer review. The article was posted July 21, 2026, and should not be used to guide clinical practice without further validation and peer review.