Probabilistic Learning

Two options are shown on each trial; the participant chooses one and receives win or no-win feedback before the next trial.

Useful in studies of impulsivity, substance use, motivation, mood, and individual differences in reward sensitivity, where trial-by-trial learning dynamics are a target outcome.

Not ideal when the study needs classical cognitive measures such as memory or attention rather than learning-rate or value-based choice data.

Enable the optional reversal point to test flexible updating as well as initial acquisition. Keep the session short enough to avoid fatigue effects on learning.

• Number of trials
• Reward probability for each arm
• Optional reversal trial index
• Feedback display duration
• Practice block enabled / disabled

• Choice proportion per arm
• Win-stay rate
• Lose-shift rate
• Post-reversal accuracy (if reversal enabled)

Win-stay and lose-shift rates describe qualitative strategy use; fitting formal RL models (e.g., Rescorla-Wagner) to the trial sequence yields learning-rate and inverse-temperature estimates that are more sensitive.

• Smartphone probabilistic-learning tasks recover meaningful individual differences in reinforcement learning parameters.
• Win-stay / lose-shift metrics are interpretable without model fitting and useful for exploratory analyses.

• Smartphone reinforcement learning feasibility review
• Two-armed bandit in mobile EMA context

• Risky Choice
• Delay Discounting
• Go/No-Go