CONTINUOUS HUMAN ESTIMATION OF A TIME-VARYING, SEQUENTIALLY DISPLAYED PROBABILITY
Abstract
This experiment examines the human ability to give a direct magnitude estimate of a time varying probability. The subject positioned a tracking lever at his estimate of the current mean of a sequentially displayed binary distribution. The distribution samples were presented at a fixed rate by two flashing lights. The distribution mean changed in step increments of varying size and spacing. The experimental variables included flash rate and a constraint on the random ness of the flash series. Detailed measures were made of both the transient and static responses to each step change. The transient response was more rapid and consistent than had been anticipated and occurred with step changes as small as 0.12. The average static response showed no systematic bias as a function of probability and had an RMS error approximately equal to that of a 17 sample average. Two simple mathematical models are derived to provide quantitative comparisons with the subjects' data. A descriptive model is also derived which satisfies some basic properties of the task behavior. The parameters for this model are selected for two specific experimental situations.
Document Details
- Document Type
- Technical Report
- Publication Date
- May 01, 1963
- Accession Number
- AD0407434
Entities
People
- Gordon H. Robinson
Organizations
- University of Michigan