Contributions of Attention to Learning in Multidimensional Reward Environments

Area of Science:

• Cognitive psychology
• Neuroscience
• Decision science

Background:

• Real-world choices involve numerous attributes, but rewards depend on a few.
• Humans integrate feature-based and conjunction-based learning for complex environments.
• The interplay between learning strategies and attention in naturalistic settings is not fully understood.

Purpose of the Study:

• Investigate how learning strategies interact to guide attention.
• Examine how attention influences future learning and choice behavior.
• Determine the role of attention in processing multidimensional reward contingencies.

Main Methods:

• Human participants (male and female) performed a three-dimensional learning task.
• Reward outcomes were predictable via an informative feature and conjunction.
• Attention-modulated reinforcement learning models were used to analyze behavior.

Main Results:

• Choice behavior and reward probability estimates best fit attention-modulated models.
• Attention was driven by the difference in integrated feature and conjunction values.
• Attention modulated learning rates but did not directly impact choice decisions.

Conclusions:

• Humans direct attention to selectively process reward-predictive attributes.
• Attention aids in creating parsimonious representations for efficient learning.
• Attention plays a crucial role in navigating complex, multidimensional environments.