Situational influences on linear and nonlinear use of information

Abstract

Comparison of linear and nonlinear models of decision-making typically leads to the conclusion that a linear model is equal to or superior to other models. In the present study, linear, conjunctive, and disjunctive models of subjects' decisions were compared in situations in which it was hypothesized that nonlinear use of information would be likely. Specifically, subjects' performance was evaluated in a 3(cue distribution: normal, positively skewed, and negatively skewed) × 2(decision-making unit: group or individual) × 3(number of cues: 2, 4, or 8) × 3(model of the decisions: linear, conjunctive, or disjunctive) design. The results indicated significantly superior subject performance in group rather than individual conditions and superiority of the linear model over the conjunctive and disjunctive models especially for group conditions and the two-cue conditions. The linear model was superior to the conjunctive and disjunctive model overall, but for many subjects the linear and conjunctive models were virtually equal. It is suggested that future studies might involve tasks and cue labels that suggest a nonlinear combination of data and that methodologies designed to identify subjects' hypotheses about functional relationships be used more frequently.