A solution to the dilemma `limiting similarity vs. limiting dissimilarity' by a method of transparent artificial intelligence

Abstract

Ecological hypotheses of ‘limiting similarity’ and ‘limiting dissimilarity’ formulate the alleged conditions for implementing the competitive exclusion. The limiting similarity hypothesis assumes that exceeding a threshold of interspecific similarity leads to competitive exclusion. The hypothesis of limiting dissimilarity states the opposite - a condition of implementing competitive exclusion is exceeding of a threshold of interspecific dissimilarity. These hypotheses clearly contradict each other and we tested them on the model of competition at different values of relative competitiveness of the species. The model of competition under study we implemented as automatic deductive inference by a method of a completely transparent and completely explainable artificial intelligence. In order to strictly verify the hypotheses, we have excluded different possible factors promoting competitive coexistence. We have achieved that by ensuring that all conditions for the studied complete competitors were equal, with the exception of competitiveness of the species. Using our individual-based cellular automata model we demonstrate that two trophically identical species aggressively competing for a single limiting resource, all other things being equal, may coexist in one limited habitat until the threshold of differences in competitiveness exceeds 31%. The obtained results reject the hypothesis of limiting similarity and confirm the hypothesis about the limiting dissimilarity. The results are of importance for evolution theory and for finding pathways to ecological sustainability. Our individual-based modelling approach, based on the transparent and explainable artificial intelligence, opens up great prospects for a variety of theoretical and applied fields.