Bounded Rationality in Queueing Systems

When a customer goes to a fast food restaurant, a café or an ATM, and has to queue for service, does he always accurately and perfectly calculate the benefits and costs of joining before making his decisions? The traditional economics and queueing literature have assumed that he does, while anecdotal evidence and experimental studies point to the contrary. Since the seminal work of Naor (1969), considerable attention has been paid to the economics of queues. Assuming all customers are fully rational, Naor (1969) shows that self-interested customers would join a more congested system than what they the social planner prescribes, and proposes levying tolls as a way to maximize social welfare. In Naor’s model, customers are assumed to be able to compute in great precision the expected utility they are about to obtain from making a decision about whether to join. Furthermore, customers are assumed to be perfectly rational and prefer joining even if the positive expected utility is negligible. One may ask, does a customer join the queueing system even if his positive expected utility is negligibly small? Are customers rational? Ariely (2009) claims that irrationality is the real invisible hand that drives human decision making. Indeed, while we are not aware of any empirical evidence that people are fully rational, there is abundant empirical evidence that people are boundedly rational. Queueing theorists had largely ignored the fact that customers are decision makers until Naor (1969). Naor (1969) and subsequent researchers following his work assumed customers are self-interested, can perfectly optimize their own utilities, and make decisions without any mistakes. Schweitzer and Cachon (2000) show that customers typically make mistakes in their decision-making in newsvendor settings. Su (2008) analyzes the problem of boundedly rational newsvendor, i.e., he is prone to errors and biases. We believe that the customers that arrive to a queuing or service system are boundedly , and in this work we study the effects and implications of this bounded rationality in a canonical queueing system.