Relating bounded rationality to fairness in macroeconomics by multi-agent deep reinforcement learning techniques