Adaptive rational expectation with forward looking agents