Costs and benefits in perceptual categorization

Area of Science:

• Cognitive Psychology
• Decision Making
• Machine Learning

Background:

• Understanding how individuals learn and adapt to environments with varying costs and benefits is crucial for decision-making research.
• Previous models often assume equal weighting of costs and benefits, which may not reflect real-world behavior.

Purpose of the Study:

• To investigate how observers learn to categorize perceptual stimuli under different cost-benefit structures.
• To determine if observers weight costs more heavily than benefits during learning.
• To examine the impact of zero versus nonzero costs on learning and performance.

Main Methods:

• Observers categorized stimuli with manipulated category costs and benefits.
• Experimental conditions included zero and nonzero costs, with equivalent optimal classifier performance.
• Nested models were applied to individual observer data across multiple trial blocks.

Main Results:

• Observer performance improved with experience, approaching but not reaching optimal levels.
• Performance was significantly worse in conditions with nonzero costs compared to zero costs.
• Data provided inconclusive evidence for the hypothesis that observers differentially weight costs over benefits.

Conclusions:

• Experience enhances cost-benefit learning, but learning asymptotes at a suboptimal level.
• The presence of costs negatively impacts performance, indicating a deviation from optimal strategy.
• Further research is needed to clarify the differential weighting of costs versus benefits in perceptual learning.