A Multi-Measure Analysis of Context Effects in Multi-Attribute Decision Making: Examining the Similarity, Attraction, and Compromise Effects

A Multi-Measure Analysis of Context Effects in Multi-Attribute Decision Making: Examining the Similarity, Attraction, and Compromise Effects Takashi Tsuzuki (tsuzuki@rikkyo.ac.jp) Department of Psychology, Rikkyo University, 1-2-26 Kitano, Niiza, Saitama 352-8558 Japan Jerome R. Busemeyer (jbusemey@indiana.edu) Department of Psychology and Brain Sciences, Indiana University 1101 E. Tenth Street, Bloomington, IN 47405 USA Abstract The similarity, attraction, and compromise effects warrant specific investigation in multi-attribute decision making. To examine these effects concurrently, we assigned 145 undergraduates to three context effect conditions. They were requested to solve 20 hypothetical purchase problems that had three alternatives described along two attribute dimensions. We measured their choices, confidence ratings, and response times. We found that adding the third alternative had significant effects for choice proportions and confidence ratings in all three conditions. The attraction effect was more prominent than the other two effects with regard to choice proportions. The compromise effect condition yielded low confidence ratings and long response times, although the choice proportion was high for the third alternative. These results indicate that the mutual relationship among choice proportions, confidence ratings, and reaction times requires theoretical investigation. Keywords: decision making; choice; context effects; similarity effect; attraction effect; compromise effect Introduction Theories of rational decision making suggest that choice is intrinsically determined by the utilities of the individual alternatives and thus unaffected by the relationships among the alternatives in the choice context. However, many studies have found violations of this tenet (Busemeyer, Barkan, & Chaturvedi, 2007; Tsetsos, Usher, & Chater, 2010). Three much-studied findings regarding such context-dependent choice effects warrant specific attention since they constitute violations of axioms fundamental to rational choice. The present paper collectively addresses these effects because they share important commonalities and can be explained using a unified framework. These findings include the attraction, similarity, and compromise effects. These effects occur with the addition of a third alternative (decoy) to the two-alternative choice set (Roe, Busemeyer, & Townsend, 2001; Tsetsos et al., 2010; Tsuzuki & Guo, 2004; Usher & McClelland, 2004). Consistent with established research, the present paper examines these effects in a two-attribute form (see Figure 1). The alternatives that constitute the core two-alternative set are commonly referred to as the target and the competitor. The target and the competitor form a trade-off—one is better than the other on one attribute, but worse on the other attribute. The third alternative is then added to this core set. Depending on the relative position of the third alternative with respect to the target, three types of phenomena are likely to occur. Two arise when the third alternative is more similar to the target than it is to the competitor. However, if a trade-off exists between the third alternative and the target, the choice probability of the target decreases relative to that of the competitor. This is called the similarity effect ( Brenner, Rottenstreich, & Sood, 1999; Tversky, 1972). In contrast, if the third alternative is inferior to the target on all attributes, the choice probability of the target should increase relative to that of the competitor. This is called the attraction effect ( Hedgcock & Rao, 2009; Huber, Payne, & Puto, 1982). The third phenomenon occurs when the third alternative rests between the target and the competitor, in which case the third alternative, now constituting a compromise between the core items, would be chosen most often. This is called the compromise effect (Mourali, Bockenholt, & Laroche, 2007; Simonson, 1989). All three of these phenomena constitute a violation of the axioms of rational choice. Numerous explanations have been provided for each of the three kinds of decoy effects (Simonson & Tversky, 1992; Tversky, 1972; Tversky & Simonson, 1993). However, Roe et al. (2001) were the first to explain all three within a single framework that was implemented in a connectionist model derived from a previous stochastic mathematical theory (Busemeyer & Townsend, 1993; Tsuzuki, Kawahara, & Kusumi, 2002). Their model (the multi-alternative decision field theory, MDFT) accounts for these findings specifically with the aid of variable lateral inhibition, which is due to similarity relations among alternatives and the momentary shifting of attention from one attribute to another. Tsetsos et al. (2010, p. 1280) remarked that “before we start, we note that these effects (the similarity, attraction, and compromise effects) were so far obtained in different studies, so until a study reports all three effects with the same