Probabilistic inference and Bayesian-like estimation in animals: Empirical evidence.

Area of Science:

• Behavioral Ecology
• Cognitive Ecology
• Evolutionary Biology

Background:

• Animals frequently encounter situations with incomplete information regarding environmental factors such as food patch quality or mate suitability.
• Theoretical models often propose that animals employ Bayesian updating to integrate prior environmental knowledge with sample data for optimal decision-making.

Purpose of the Study:

• To investigate empirical evidence supporting the hypothesis that animals make decisions in a Bayesian-like manner under uncertainty.
• To examine how animals' sensitivity to variance, probabilistic inferences, and behavioral responses align with Bayesian model predictions.

Main Methods:

• Reviewing studies on animal decision-making that assess sensitivity to frequency distribution variances.
• Analyzing research where animals make probabilistic inferences based on relative frequencies of preferred and non-preferred resources.
• Comparing observed animal behaviors with predictions from Bayesian models in two distinct experimental approaches.

Main Results:

• Nine bird and mammal species exhibited patch exploitation behaviors consistent with Bayesian model predictions.
• The decision-making processes of three bird species, bumblebees, and fruit flies aligned with Bayesian principles when learning resource distributions.
• Animals demonstrated sensitivity to variance and made probabilistic inferences that maximized reward likelihood.

Conclusions:

• Empirical evidence suggests animals may utilize a Bayesian-like approach to combine prior and sample information for decisions under uncertainty.
• While findings support the generality of Bayesian-like decision-making, further research across diverse species is necessary.
• This study highlights the cognitive mechanisms animals may employ to navigate complex environments and optimize fitness.