Endogenous Strategy in Exploration

    loading  Checking for direct PDF access through Ovid

Abstract

We examined the characteristics of endogenous exploratory behaviors in a generalized search task in which guidance signals (e.g., landmarks, semantics, visual saliency, layout) were limited or precluded. Individuals looked for the highest valued cell in an array and were scored on the quality of the best value they could find. Exploration was guided only by the cells that had been previously examined, and the value of this guidance was manipulated by adjusting spatial autocorrelation to produce relatively smooth and rough landscapes—that is, arrays in which nearby cells had unrelated values (low correlation = rough) or similar values (high correlation = smooth). For search in increasingly rough as compared with smooth arrays, we found reduced performance despite increased sampling and increased time spent searching after revelation of a searcher's best cell. Spatially, sampling strategies tended toward more excursive, branching, and space-filling patterns as correlation decreased. Using a novel generalized-recurrence analysis, we report that these patterns reflect an increase in systematic search paths, characterized by regularized sweeps with localized infilling. These tendencies were likewise enhanced for high-performance as compared with low-performance participants. The results suggest a trade-off between guidance (in smooth arrays) and systematicity (in rough arrays), and they provide insight into the particular strategic approaches adopted by searchers when exogenous guiding information is minimized.

Related Topics

    loading  Loading Related Articles