Using Machine Learning to Understand and Influence Human

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . x 1 Classification As Model of Human Categorization . . . . . . . 1 1.1 Review of Classification in Machine Learning and Cognitive Psychology . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 1.2 Semi-Supervised Learning Assumptions . . . . . . . . . . . 5 1.3 Translating Between ML and CP . . . . . . . . . . . . . . . 6 2 Semi-Supervised Models of Human Categorization Behavior 9 2.1 Exemplar Model as Kernel Density Estimation . . . . . . . 9 2.2 Prototype Model as Mixture of Gaussians . . . . . . . . . . 12 2.3 Rational Model as Dirichlet Process Mixture Model . . . . 16 3 Semi-Supervised Effects Due to Distribution of Unlabeled Data: Previous Evidence . . . . . . . . . . . . . . . . . . . . . . . . . . 22 3.1 Experiment 1: SSL Distribution Effects . . . . . . . . . . . . 22 3.2 Experiment 2: Social Categories . . . . . . . . . . . . . . . . 25 4 Semi-Supervised Effects Due to Order of Unlabeled Data . . . 30 4.1 Human Experiment . . . . . . . . . . . . . . . . . . . . . . . 31 4.2 Model Comparison . . . . . . . . . . . . . . . . . . . . . . . 33 5 What Parameters Are Affected in Semi-Supervised Effects? . . 38 5.1 Competing Hypotheses . . . . . . . . . . . . . . . . . . . . . 38