Bidirectional associative memory networks applied to modeling non-neoclassical economic behaviour

Artificial neural networks can be applied to analyzing and understanding economic behaviour. Feedback networks, such as the bidirectional associative memory (BAM) network, are appropriate when economic agents make decisions based on other agents´ behavior. In the case presented, the BAM weight matrix represents the influence of the group of workers on any one worker. Each X vector represents a specific worker´s characteristics and Y vector represents the results given the firm´s work rules. It is shown that imposing more constraints on the workers polarized them into two extreme performance groups with an overall result of reducing the effort offered by poorer workers. The presence of poor workers causes good workers to work harder. It is not conclusive that replacing the poor workers with better workers increases the productivity of the group