Measuring Social Motives: The Reliability and Validity of Parametric Estimates Derived from Dictator Game Choices.
Three experiments were conducted; all measured social motives with DGs, followed-up by measures of constructs relevant for validation. Data were analyzed via Bayesian simultaneous equation procedures where motives were modeled as (latent) parameters in a multidimensional IRT-model, estimated jointly with relevant validation parameters. Reliability was assessed via population-, person-, and item-level estimates of precision. Internal validity was investigated by comparing the fit of the assumed utility model against the fit of alternative models; content validity was investigated via a two-stage procedure where DG subjects rated previously answered items on relevant characteristics; construct validity was investigated by relating motives to plausible correlates such as self-reported altruism, psychopathy, empathy, generalized trust, and risk-preference.
Results showed adequate precision in population-level estimates of motives, but exploratory analyses revealed (not surprisingly) that individual-level precision in particular was highly influenced by sample-size; this warrants further investigation on how to determine the optimal within- and between-subject samples size for a desired level of precision. Results support the internal validity of the utility model, as well as the content validity of DG-items. Modest correlations between motives and expected correlates were found, which establishes construct validity. Although these findings are promising, caution is advised in interpreting social motives as personality traits. First, the stability of social motives needs additional investigation; second, exploratory analyses suggest that both individual and situational factors mediate the expression of social motives; thus the etiology of social motives in general needs further elucidation.