Citation

Wickens CD, Clegg BA, Witt JK, Smith CAP, Herdener N, Spahr KS. Theor. Issues Ergonomics Sci. 2020; 21(2): 220-238.

Copyright

(Copyright © 2020, Informa - Taylor and Francis Group)

DOI

10.1080/1463922X.2019.1679907

PMID

unavailable

Abstract

Understanding the variability of trends and other continuously distributed quantities is a vital ability underlying many safety critical decisions, such as how widely to search for a downed aircraft, or whether to prepare for evacuation in the face of an uncertain hurricane or hurricane track. We first review the sparse research on this topic which indicates a general systematic tendency to underestimate such variability, akin to overconfidence in the precision of prediction. However, the magnitude of such underestimation varies across experiments and research paradigms. Based on these existing findings, and other known biases and vulnerabilities of the perception and cognition of multiple instances, we define the core elements of a computational model that can itself predict three measures of performance in variability estimation: bias (to over or underestimate variability), sensitivity (to variability differences) and precision (of variability judgements). Factors and approximate weighting in influencing these measures are then identified regarding attention, the number of instances across whose variability is estimated, the time delay affecting the memory system employed, familiarity of material, the anchoring heuristic and the method of judgement. These are then incorporated into foundations for a linear additive model.

Language: en

Keywords

bias; decision making; judgement; model; Prediction; trajectories; variability