Estimating variability across numeric and spatial information
Date
2020
Authors
Spahr, Kimberly S., author
Clegg, Benjamin A., advisor
Wickens, Christopher D., advisor
Prince, Mark, committee member
Smith, Charles, committee member
Journal Title
Journal ISSN
Volume Title
Abstract
Research has demonstrated the difficulty of estimation and prediction, particularly in complex and uncertain conditions. Specifically, humans lack precision or are biased in making estimates of variability from continuously distributed stimuli, such as hurricane trajectories (spatial information) or sets of random numbers (numeric information). Conversely, people tend to provide calibrated estimates of average behavior for both spatial and numeric stimuli. Using either spatial or numeric stimuli, past studies noted that people tend to underestimate variability but provide accurate mean estimates. Nonetheless, no experiments have utilized both spatial and numeric stimuli to identify the extent to which people estimate variability, and to a lesser extent, mean behavior, across different types of information. This individual differences perspective holds significant implications for training and support in making calibrated decisions under uncertainty. The current study addressed this gap by presenting participants with a spatial task and a numeric task, each of which assessed knowledge and calibration to variability and means. Using cross-task correlational analyses, this study explored the extent to which similar mechanisms might underlie performance in both domains of stimuli. During the spatial task, participants learned the location of varying trajectories, and then reported on the likelihood of possible trajectory endpoints (spatial variability) and the average trajectory. During the numeric task, participants viewed lists of random numbers, and estimated the mean and spread of these lists (numeric variability). A correlational analysis revealed that participants who gave more accurate estimates of variability on the spatial task were not necessarily more accurate when estimating numeric variability. Such findings indicate that different cognitive processes likely support the understanding of variability for different types of information. Additional research is necessary to elucidate which cognitive mechanisms are involved; possible systems include working memory and numeracy. Participants expressed a similar overestimation bias to variability across both tasks. This bias trend does not replicate prior literature for either spatial or numeric information, and future studies will focus on how to induce participants to change their response biases. Finally, mean estimation performance correlated across tasks, meaning that those who were more accurate when estimating spatial means were more likely to accurately estimate numeric means.
Description
Rights Access
Subject
human performance
variability
individual differences
cognition