Stojic, Hrvoje; Henrik Olsson and Pantelis P. Analysis

Choosing between options characterized by multiple cues can be a daunting task. Do people integrate all information when making choices or do they employ lexicographic heuristics that ignore some of it? Notably, integrative strategies require knowing exact cue weights, whereas lexicographic heuristics can operate by merely knowing the cue importance ranks. Here we investigate how people employing these strategies learn about the structure of the environment. We introduce the cue-learning while choosing paradigm, where participants choose between three options whose cues are linearly related to criterion values. Through feedback about the chosen option they could learn how cues were related to criterion values. We uncovered the evolution of participants’ knowledge by fitting a least-mean-squares learning model jointly to the experienced options and to predictions in an estimation task that they completed at the end, and validated these results by asking people to periodically estimate cue importance. We pitted the integrative Weighted-additive (WADD) strategy against ∆-Elimination, a novel lexicographic heuristic, in capturing participants’ choices. While the WADD strategy accounted for most of the participants’ choices, the ∆-Elimination heuristic captured a sizable group of participants. Regardless of the strategy they employed, participants learned precise cue weights. Our results suggest that cue-weight learning precedes the strategy selection and that the two processes are more independent than previously thought. They also answer the long standing riddle of how people using lexicographic heuristic strategies learn cue ranks.