A NEURAL-NETWORK MODEL FOR ATTRIBUTE-BASED DECISION-PROCESSES

We propose a neural model of multiattribute-decision processes, based on an attractor neural network with dynamic thresholds. The model may be viewed as a generalization of the elimination by aspects model, whe reby simultaneous selection of several aspects is allowed. Depending o n the amount of synoptic inhibition, various kinds of scanning strateg ies may be performed, leading in some cases to vacillations among the alternatives. The model predicts that decisions of a longer time durat ion exhibit a lower violation of the simple scalability low, as oppose d to shorter decisions. Furthermore, the model is suggested as a gener al attribute-based decision module. Accordingly, various decision stra tegies are manifested depending on the module's parameters.