A Scoring Rule to Train Probability Assessors.
Abstract
Two experiments showed that a test based on a propert scoring rule could improve probability assessments. The test contained three sessions of factual questions. The Ss indicated which of two answers was more likely to be correct for each question and then used odds assessments as an indication of how sure they were about each answer. The first and third sessions used the same set of 50 Questions; Ss were not informed whether their answers were correct. The second session used a different set of 75 questions; Ss were informed whether each answer was correct and, in addition, received a score based upon a proper scoring rule (a function that increases the score earned when higher odds are assigned to the correct answer). One experiment used college students and the other used intelligence analysts. In both cases most Ss' earned higher scores in the third than in the first session, presumably as a result of experience with the scoring rule during the second session. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 30, 1972
- Accession Number
- AD0751562
Entities
People
- Cameron R. Peterson
- Jane Hoffman
Organizations
- University of Michigan