This paper presents MADeM, a multi-modal agent decision making to provide virtual agents with socially acceptable decisions. We consider multi-modal decisions as those that are able to merge multiple information sources received from a MAS. MADeM performs social decisions since it relies on auctions, a well known market-based coordination mechanism. Our social agents express their preferences for the different solutions considered for a specific decision problem, using utility functions. Therefore, coordinated social behaviors such as task passing or planned meetings can be evaluated to finally obtain socially acceptable behaviors. Additionally, MADeM is able to simulate different kinds of societies (e.g. elitist, utilitarian, etc), as well as social attitudes of their members such as, egoism, altruism, indifference or reciprocity. MADeM agents have been successfully verified in a 3D dynamic environment while simulating a virtual university bar, where different types of waiters (eg. c...