On-line Crowdsourcing of Multi-Timescale Inference Strategies

Summary

The investigators will record behavioral responses from human participants on crowdsourcing platform Prolific to identify the decision strategies humans apply short-term, long-term, and multi timescale (across both timescales) inference tasks. Participants will perform a Jar Switching Task (described in Research Strategy Aim 1- "Jar Switching Task") in which balls are drawn with replacement from one of two known jars. The current jar in use switches based on an underlying change rate across trials. Subjects must perform three task blocks: 1) report the current jar in use (short-term inference), 2) predict the subsequent jar (long-term inference), and 3) both report and predict the jars (multi-timescale inference). The investigators will record these responses, the number (and sequence) of balls drawn, and the response time (time from the end of trial until the response) for each trial. Subjects will perform all blocks (parameters and number of blocks to be determined by inference model development and testing prior to task development) so that we can compare responses at each timescale. Since participants participate voluntarily for small sums of money (around $10/ hour based on duration of task) and the investigators' previous studies have collected over 200 subjects in a matter of days, they will aim to record behavioral data from 1000 subjects. This number allows them to address the broad range of subject variability expected using Bayesian statistical methods such as Bayes factors.

Eligibility

Age range

18+ years

Sex

All

Healthy volunteers

Yes

Interventions

• Behavioral
  psychophysics

  Subjects will be shown a pair of jars each filled with the same absolute number but different ratios of red/blue balls. Three blocks of trials exist, short-term inference (subjects must identify which of the equally likely jars is currently in use), long-term inference (subjects must predict the subsequent jar in use on the next trial, where jar switches are correlated across trials), or multi-timescale inference (subjects must both report the current jar and predict the subsequent jar, combining the short and longterm inferences). Subjects will draw balls as many balls as they wish with replacement (free response) during the short-term and multi-timescale blocks but will know the current jar in use for long-term inference.

Location