Stimulus discriminability may bias value-based probabilistic learning.

Area of Science:

• Cognitive Neuroscience
• Behavioral Economics
• Psychology

Background:

• Reinforcement learning tasks assess learning from positive and negative outcomes.
• Comparing learning across tasks requires accounting for stimulus differences.
• Perceptual factors like stimulus discriminability can influence task performance.

Purpose of the Study:

• To investigate the impact of stimulus discriminability on reinforcement learning.
• To determine if perceptual factors bias assessments of learning from positive versus negative outcomes.
• To establish the necessity of controlling for stimulus discriminability in reinforcement learning research.

Main Methods:

• Two versions of a probabilistic learning task using Hiragana characters with varying reward probabilities.
• A separate discrimination experiment to assess character perceptual discriminability.
• A large-scale web-based experiment to replicate and extend findings.

Main Results:

• Task performance was significantly influenced by the version of the task.
• Perceptual discriminability of stimuli affected participants' ability to learn reward probabilities.
• Differences in learning curves and reward sensitivity were observed based on stimulus discriminability.
• Performance biases in assessing learning from positive versus negative outcomes were linked to stimulus discriminability.

Conclusions:

• Perceptual factors play a critical role in guiding reinforcement learning.
• Stimulus discriminability must be controlled to accurately infer individual differences in reinforcement learning.
• Future research should consider stimulus properties when designing and interpreting reinforcement learning experiments.