Biological systems simultaneously maintain high fidelity of information transmission and actively generate variability. Close to this work lie studies of mutation modifiers, stress-induced mutagenesis, and stochastic switching in changing environments. In those studies, however, passive errors and regulated instability are usually not separated as distinct sources of uncertainty with different structure in state space.
A minimal replicator model was built from two subunits, A and B, where the loss of B destroys the functional complex but opens access to a more efficient AAB complex. The model was then generalized to the reversible turnover of cellular factors and to a general transition kernel that separates passive error from structured active destabilization. The analysis used standard Shannon quantities: state entropy, temporal mutual information, and Kullback-Leibler divergence. In addition, published yeast data were reanalyzed: experiments on controlled phenotypic switching and measurements of adaptive mutants in static and fluctuating environments.
The central result is called the complementarity of stabilization and destabilization: lowering passive error always increases the marginal value of active destabilization, which means selection can simultaneously boost accuracy and sustain regulated variability. In the high-accuracy limit, optimal active destabilization remains positive. The aging model shows that when selection on delayed consequences weakens, optimal maintenance declines, whereas active destabilization persists or grows. In the sexual model, shared genetic control under different optima in females and males generates intralocus sexual conflict and a measurable load. The empirical analysis confirms that rapid switching pays off when the environment changes quickly, and that the dependence of fitness on transition history grows with the divergence of static environments.
Unlike neighboring studies, where instability is treated as a single scalar parameter, this work introduces a distinction between unstructured error and structured exploration of state space. That makes it possible to explain why high accuracy and active variability do not compete but complement each other. The work links information theory to antagonistic pleiotropy and offers a common language for describing molecular, cellular, and organismal transitions.
The model can be extended to explicit transition matrices, a time-varying probability of environmental change, density dependence, and age-specific demographic schedules. It would also be promising to include epigenetic reprogramming and sex regulation as mechanisms for resolving conflict.
The results apply to evolutionary biology, gerontology, epigenetics, and the study of sexual selection.
The next step is to test the full predicted pattern in a single biological system: a low background error rate, regulated active variability, a direct benefit, a cost, and delayed consequences that depend on age and sex.
The work directly links selectively advantageous instability to the unsolved problem of the evolution of aging: why mechanisms that confer early benefits persist despite late-life costs. It also formalizes intralocus sexual conflict as a divergence in the optimal rates of information destabilization.
In the replicator model, losing subunit B destroys the working complex, yet that very loss opens the path to a more efficient AAB complex: destroying part of a whole can be the condition for improvement.
The idea that controlled forgetting and regulated variability can be useful for adaptation echoes the concept of "forgetting" as a resource in Stanisław Lem's novel His Master's Voice.
Not published on arXiv. Here are its page, the PDF and the formalization: you can check it yourself, not only read our account.