The role of first impression in operant learning

Area of Science:

• Cognitive Psychology
• Computational Neuroscience
• Behavioral Economics

Background:

• Operant learning involves acquiring behaviors through consequences.
• Understanding the computational basis of decision-making is crucial.
• The impact of initial experiences on learning is not fully understood.

Purpose of the Study:

• To quantify the lasting effect of first experiences on behavior in operant learning.
• To investigate the computational principles underlying this 'outcome primacy' effect.
• To model behavior using reinforcement learning and explore the role of initial conditions.

Main Methods:

• Analysis of over 200,000 choices in a repeated-choice experiment.
• Application of a standard, model-free reinforcement learning algorithm.
• Development of a novel nonparametric method to characterize action-selection rules.

Main Results:

• First experience outcomes have a substantial and lasting impact on subsequent behavior (outcome primacy).
• Outcome primacy explains the underweighting of rare events and underestimation of small probabilities.
• A reinforcement learning model where first experience resets action values accurately predicts behavior, outperforming other models.

Conclusions:

• First experience exerts a disproportionately large influence on actions, akin to primacy effects in other cognitive domains.
• A resetting mechanism for initial conditions in reinforcement learning underlies outcome primacy.
• This resetting mechanism may explain other forms of primacy in cognitive psychology.