Apparent statistical inference in crows may reflect simple reinforcement learning

Area of Science:

• Cognitive ethology
• Animal behavior
• Decision-making under uncertainty

Background:

• Crows exhibit complex decision-making behaviors.
• Previous research suggested crows use statistical inference for choices.
• Johnston et al. (2023) proposed an analogue magnitude system for this inference.

Purpose of the Study:

• To investigate the mechanisms underlying crows' choice behavior.
• To test whether crows engage in statistical inference or simpler learning processes.
• To challenge the interpretation of crows' behavior as statistical inference.

Main Methods:

• Training crows to associate stimuli with varying reward probabilities (10%-90%).
• Conducting choice tests between paired stimuli.
• Comparing empirical findings with predictions from a reinforcement learning model versus statistical inference models.

Main Results:

• Crows consistently preferred stimuli with higher reward probabilities.
• Preference magnitude was influenced by the absolute difference, not the ratio, of probabilities.
• A simple reinforcement learning model accurately reproduced key behavioral patterns.

Conclusions:

• Crows' decision-making can be explained by reinforcement learning, not necessarily statistical inference.
• The results challenge the necessity of an analogue magnitude system for representing abstract probabilities.
• Behavioral patterns do not require complex retrieval and application of calculated reward probabilities.