Algorithmic game theory has made great strides in recent decades by assuming standard economic models of rational agent behavior to study outcomes in distributed computational settings. From the analysis of Internet routing to the design of advertisement auctions and crowdsourcing tasks, researchers leveraged these models to characterize the performance of the underlying systems and guide practitioners in their optimization. These models have tractable mathematical formulations and broadly applicable conclusions that drive their success, but they rely strongly on the assumption of rationality.
The assumption of rationality is at times questionable, particularly in systems in which human actors make most of the decisions and in systems that evolve over time. Humans, simply put, are bad at thinking about the impact of their actions on their environment and their future. We see this every day in the way we manage our time. Students cram for exams despite planning not to, and even though it is well documented that well-spaced studying produces improved learning results with equal effort. Humans in lab experiments also consistently exhibit similar irrational time-inconsistent planning and procrastination behavior.
No entries found