A Quantitative Test Of Computational Models Of Multialternative Context Effects

When people choose between multiple alternatives it has been observed that the probability of choosing one option over another often depends on the additional options available. For instance, one's preference for a high-end notebook computer over a cheaper midrange model can reverse when a third, low price, but bulkier, computer is also considered. Importantly, such context effects violate fundamental choice axioms of most standard economic choice theories. To account for these effects various computational models have been proposed. The present work examines two prominent models that have been proposed to simultaneously account for the similarity, compromise, and attraction effects: multialternative decision field theory (MDFT; Roe, Busemeyer, & Townsend, 2001) and the multiattribute linear ballistic accumulator (MLBA; Trueblood, Brown, & Heathcote, 2014). Because the models imply quite different psychological processes, we focus our comparison on the various mechanisms the models use to produce each effect. We test the models using data from a large within-subjects consumer choice experiment in which all three effects occurred. Overall, we find that both MLBA and MDFT have valuable mechanisms for understanding multialternative context effects. When considering each context effect in isolation MDFT gave the better account of the similarity and attraction effects, while MLBA better described the compromise effect. We conclude that three mechanisms-attention switching and distance-dependent inhibition from MDFT and extremeness aversion from MLBA-appear crucial. These results provide insights into the psychological mechanisms underlying human decision making and should be considered in further development of decision theories.