Boitnott, Joshua Forrest - Applications of Individual Evolutionary Learning...

View the thesis

This thesis has been submitted to the Library for purposes of graduation, but needs to be audited for technical details related to publication in order to be approved for inclusion in the Library collection.
Term: 
Spring 2017
Degree: 
Ph.D.
Degree type: 
Thesis
Department: 
Department of Economics
Faculty: 
Arts & Social Sciences
Senior supervisor: 
Jasmina Arifovic
Thesis title: 
Applications of Individual Evolutionary Learning
Given Names: 
Joshua Forrest
Surname: 
Boitnott
Abstract: 
This research investigates three applications of the Individual Evolutionary Learning (IEL) model. Chapter 2 utilizes a horse-race approach to investigate the overall performance of 4 learning algorithms in games with congestion. The games utilized are Market Entry games and Choice of Route games. I show that a version of the IEL has the best fit of the experimental data relative when the experimental subjects have full information. Chapter 3 (joint work the Jasmina Arifovic and John Duffy) applies the IEL to games with correlated equilibrium suggested by an external third party. The IEL nearly perfectly matches the behavior of experimental subjects playing the Battle of the Sexes game, but requires an adjustment to the initial conditions to match the behavior of experimental subjects in the Chicken game. Chapter 4 extends the Individual Evolutionary Learning with Other-Regarding Preferences (IELORP*) model to force the algorithm to match the discrete nature of the experimental choices and introduce beliefs via adaptive expectations. The algorithm continues to match the stylized facts associated with the standard LPGG, but does not appear to extend to games where beliefs are elicited using monetary incentives.
Keywords: 
Agent-Based Modeling; Evolutionary Algorithm; Experimental Economics; Learning
Total pages: 
139