Modeling the animal in its world and the canonical problem for habitat selection
Marc Mangel · Israel Journal of Ecology and Evolution · 2025
Abstract Behavioral and evolutionary ecology made enormous progress in the last 50 years by using the assumption that the modeler/observer is external to the organism and its environment. This allows specifying details of the environment (e.g. predation risk or the probability of finding food) and then using a fitness optimization model to predict the behaviors that are the end point of natural selection. Doing so can be called the third-person perspective of the organism. More than 80 years ago, Jakob von Uexküll argued that a first-person perspective is possible if we replace the external observer’s description of the environment by the organism’s subjective characterization of itself and its surroundings based on its sensory data, and allow those sensory data to shape behavior. The first-person umwelt model becomes an evolutionary one when the genes determining the sensory responses evolve. I use a canonical problem of habitat selection (which will always be an important problem in biology and which was one of Leon Blaustein’s favorite topics of research) to illustrate construction of first-person umwelt and third-person fitness optimization models. A canonical problem is the simplest but still interesting form of a collection of similar problems, with a focus on what is essential to the collection as a whole. In particular, a canonical problem does not model any particular situation, but has much in common with many situations. Here, the canonical problem focuses on an organism that needs to find a refuge from a harsh environmental season that begins at a fixed time when refuges vary in the level of protection from the harsh environment. Individuals who survive the harsh environmental season successfully reproduce. When searching for refuges, organisms experience predation risk so that the third-person fitness optimization model answers “when an organism encounters a habitat of a specific quality at a given time, is it predicted to settle or continue searching”? The first-person umwelt model is based on the assumptions that i) the organism has sensory inputs that provide information on quality of a habitat (settling is more likely in higher quality habitats) and the remaining time before the onset of the harsh environmental season (settling is more likely when less time remains), ii) the sensory information is combined to determine behavior, and iii) the genotypic architecture underlying the sensory functions evolves by natural selection. After developing predictions from the models by simulating populations following the rules developed from each, I use effect size measured by Cohen’s d to explore how evolution of the genes of the response functions in the first-person umwelt model affects survival and convergence of the predictions of the first-person umwelt and third-person fitness optimization models.