Mice adaptively generate choice variability in a deterministic task.

Area of Science:

• Neuroscience
• Behavioral Economics
• Computational Biology

Background:

• Decision-making processes are often assumed to be influenced by external stochasticity.
• The intrinsic capacity for generating variability in choice behavior remains an open question.

Purpose of the Study:

• To investigate whether decision-makers can intrinsically generate variability in their choices.
• To model animal decision-making using reinforcement learning in a controlled, deterministic setting.

Main Methods:

• Mice were trained in a deterministic environment rewarding non-repetitive choice sequences.
• Reinforcement learning models were used to analyze the decision-making strategies.
• Behavioral responses were analyzed under varying reward delivery conditions.

Main Results:

• Mice progressively increased their choice variability over time.
• Animals adopted a pseudo-random selection strategy, deviating from theoretically optimal, more rewarding sequences.
• Mice exhibited blindness to changes in temporal reward structure after establishing random choices.

Conclusions:

• Decision-making can self-generate variability and randomness, independent of environmental stochasticity.
• Intrinsic variability may serve as a strategy to ensure a high success rate in certain decision tasks.
• Animal behavior demonstrates a capacity for generating randomness even in predictable reward scenarios.