Download now Free registration required
The author investigates when payoff can be beneficial in a learning process of a boundedly rational, risk averse agent. Destroying payoff reduces average returns but can - as a "Self punishment" - accelerate learning. This smoothness payoffs and can thereby increase expected utility. If we think of human decision making as "Boundedly rational" in the sense of Simon, then a boundedly rational agent may be well aware of his or her cognitive limitation in deciding in every situation which alternative is the best.
- Format: PDF
- Size: 190.81 KB