Neural voting machines

In theories of cognition that view the mind as a system of interacting agents, there must be mechanisms for aggregate decision-making, such as voting. Here we show that certain voting procedures studied by social scientists can be implemented as recurrent neural networks. For example, a standard "winner-take-all" network can determine which of a number of competing alternatives garners a plurality of votes. Similarly, in the special case where voters share a model governing the different rankings of alternatives, the Borda procedure can easily be computed. In the face of voter un-certainties, this Borda network returns the maximum likelihood choice.

• Type: Journal
• Year: 2006
• Where: NN
• Authors: Whitman Richards, H. Sebastian Seung, Galen Pickard
• Keywords: Aggregate Decision-making, Certain Voting Procedures, Neural Networks, NN 2006, Recurrent Neural Networks