Judgment under Chaos

Abstract

This study extends the experimental paradigm for studying the performance of human subjects in short-range predictions of chaotic sequences, and reframes the original research question posed by Neuringer and Voss (1993) so that it is informed by the literature on judgment under stochastic uncertainty. Unlike their study or its replication and extension by Metzger and Theisz (1994), in this study, subjects' performances are compared under identically distributed chaotic and random sequences. Moreover, both persistent and anti-persistent chaotic sequences are employed to investigate whether subjects' prediction patterns are influenced by persistence. Lastly, in addition to simple accuracy measures, several other performative indicators are employed as suggested by judgment research and the mathematics of chaos. The main differences that emerge between judgments under randomicity and judgments under chaos are greater accuracy under chaos (thereby replicating earlier findings) and less under-dispersion in the restricted sense of correlated prediction errors. Somewhat more surprising are the similarities, particularly the high accuracy with which subjects approximate the (anti-) persistence of both random and chaotic sequences. The discussion outlines some implications for our understanding of judgment under uncertainty and human behavior in chaotic environments.