This paper presents a theory of multi-alternative, multi-attribute preferential choice. It is assumed that the associations between an attribute and an available alternative impact the attribute's accessibility. The values of highly accessible attributes are more likely to be aggregated into preferences. Altering the choice task by adding new alternatives or by increasing the salience of preexisting alternatives can change the accessibility of the underlying attributes and subsequently bias choice. This mechanism is formalized by use of a preference accumulation decision process, embedded in a feed-forward neural network. The resulting model provides a unitary explanation for a large range of choice-set-dependent behaviors, including context effects, alignability effects, and less is more effects. The model also generates a gain-loss asymmetry relative to the reference point, without explicit loss aversion. This asymmetry accounts for all of the reference-dependent anomalies explained by loss aversion, as well as reference-dependent phenomena not captured by loss aversion.
PsycINFO Database Record (c) 2013 APA, all rights reserved.