Freely foraging macaques value information in ambiguous terrains
Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Among non-human primates, macaques are recognized for thriving in a wide range of novel environments. Previous studies show macaque’s affinity for new information. However, little is known about how information-seeking manifests in their spatial navigation pattern in ambiguous foraging terrains, where the location and distribution of the food are unknown. We investigated the spatial pattern of foraging in free-moving macaques in an ambiguous terrain, lacking sensory cues about the reward distribution. Rewards were hidden in a uniform grid of woodchip piles spread over a 15 sqm open terrain and spatially distributed according to different patchy distributions. We observed Lévy-like random walks in macaques’ spatial search pattern, balancing relocation effort with exploration. Encountering rewards altered the foraging path to favor the vicinity of discovered rewards temporarily, without preventing longer-distance travels. These results point toward continuous exploration, suggesting that explicit information-seeking is a part of macaques’ foraging strategy. We further quantified the role of information seeking using a kernel-based model, combining a map of ambiguity, promoting information seeking, with a map of discovered rewards and a map of proximity. Fitting this model to the foraging paths of our macaques revealed individual differences in their relative preference for information, reward, or proximity. The model predicted that a balanced contribution of all three factors performs and adapts to an ambiguous terrain with semi-scattered rewards, a prediction we confirmed using further experimental evidence. We postulate an explicit role for seeking information as a valuable entity to reduce ambiguity in macaques’ foraging strategies, suggesting an ecologically valid way of foraging ambiguous terrains.
Graphical Summary
A) The experimental setup. Monkey Vin in a floor foraging session in the open arena of the Exploration Room. From a monkey’s point of view, the terrain appeared as a uniform grid of woodchip piles arranged so that he could walk in the gap between the piles. Therefore, the terrain was ambiguous in the sense that it did not provide any sensory cue about the distribution and location of hidden rewards. Inset : Example foraging paths from one experimental session. Less than 20% of the woodchip piles, arranged in a disk shape on the floor, hid reward pieces (full piles; purple dots), and the rest were empty (black dots) B-D) Results of statistical analysis on the experimental data (top) and a generative or fitted kernel-based model of spatial foraging (bottom). B) Top: The distribution of step sizes, defined as distances between consecutive pile searches, is linear with a negative slope, suggesting a Lévy-like distribution. A Lévy-like distribution suggests that the monkey balanced energy preservation with exploration by taking short steps most of the time, medium-length steps sometimes, and long steps rarely. Bottom: A similar distribution of step sizes was produced in simulated foraging sessions in which the foraging agent makes choices among 108 locations, factoring in reward gain, information gain, and proximity of the pile. C) Top: The average of the step size immediately after encountering filled piles was shorter than that of empty piles. Bottom: similar results are achieved using the simulated agent in panel B bottom. D) Top: Two types of foraging terrains: a localized map with the disk-shaped arrangement of filled piles and a scattered map with 4 clusters of 3 filled piles. Bottom: weights of information seeking in one monkey’s foraging choices, when the kernel-based model was fit to experimental data, show a more prominent role for information seeking on scattered terrains.
