SEQUENTIAL INFORMATION SEEKING: EFFECTS OF THE NUMBER OF TERMINAL ACTS AND PRIOR INFORMATION
Abstract
An experiment was conducted by means of a digital computer in which 54 human Ss were faced with the task of sampling from a hypothetical binomial universe in which a proportion, p, of all observations were 'top quality.' Ss sampled the universe sequentially, stopping after some number of observations had been made to make a terminal decision by selecting the one of the S's mutually exclusive and exhaustive subsets of the unit interval which S believed contained p. The four experimental treatments were defined by the four combinations of the two decision partitions of the unit interval, one involving 3 possible terminal acts, the other having 5 alternatives, and the two prior frequency distributions, one a rectangular distribution over (0 - 1), the other being negatively skewed. Analysis of variance of the number of predecision observations taken indicated a) significant individual differences; b) significant S by treatment interactions; c) differences attributable to the decision partitions with more observations being taken in the 5-act case than in the 3-act case; and d) no effect of prior frequency distributions, but a tendency to take more observations in the second 16 trials than in the first 16.
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 01, 1964
- Accession Number
- AD0614708
Entities
People
- David M. Messick
Organizations
- University of North Carolina at Chapel Hill