Human Forage for Reward: Multi-Experiment Bandit Task Data from 802 Participants

802 participants performed online and in-person restless bandit tasks across five experiments to study human foraging behavior in reinforcement learning contexts. The dataset, created by Meriam Zid and last updated in May 2026, includes behavioral data and model fits for tasks varying in volatility, option linkage, and the number of alternatives. Data is provided in formats for Python and MATLAB analysis.

Use Cases

• Compare traditional reinforcement learning models to foraging-RL models based on the provided model fits.
• Analyze how environmental volatility influences choice behavior based on the variable hazard rates and step sizes described for Experiment 3.
• Study the effect of option linkage on decision-making using the modified two-armed bandit task from Experiment 2.
• Investigate how the number of available options (2, 3, or 4) affects exploration strategies based on the design of Experiment 4.

Strengths

• Includes data from 802 total participants across five distinct experiments.
• Provides pre-computed model fits for traditional RL, foraging-RL, and Win-Stay-Lose-Shift models.
• Data is formatted for analysis in both Python and MATLAB, as indicated by the file descriptions.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment for certain analyses.
• The description metadata is limited; actual data structure and quality require manual inspection after download.

Provenance

Source

figshare

Collection Method

Data collected via online experiments on Amazon Mechanical Turk and one in-person experiment.

Freshness

Last updated 2026-05-06 21:17:41; freshness should be verified.