-
Introduction
Part I: Developmental Psychology
-
Android Epistemology for Babies
-
Introduction
-
Children
-
The Platonic Theory of Cognitive Development
-
The Theory Theory
-
Android Epistemology
-
Issues
-
Another Way for Nerds to Make Babies: The Frame Problem and Bayes Nets in Developmental Psychology
-
The Frame Problem
-
A Toy Introduction to the Markov Assumption
-
The Causal Markov Assumption
-
Causal Bayes Nets
-
The Utility of Causal Bayes Nets
-
Heuristics and Concept Formation
-
Experiments
-
Conclusion
Part 2: Adult Judgments of Causation
-
A Puzzling Experiment
-
The Baker Experiment
-
The Baker Interpretation, Allen’s Criterion of Rationality and the Rescorla-Wagner Model
-
Spellman’s Interpretation and Its Difficulties
-
The Problem for Psychologists
-
The Puzzle Solved
-
Marilyn Vos Savant Meets Rescorla-Wagner: A Crucial Experiment
-
Introduction
-
Conditional Association and the Monte Hall Game
-
Testing the Rescorla-Wagner Model
-
Cheng Models
-
Introduction
-
Cheng’s Model of Human Judgement of Generative Causal Power
-
Preventive Causes
-
Generative Interaction
-
Other Forms of Interaction
-
Cheng Models as Bayes Nets
-
Estimating the Causal Graph
-
Conclusion
-
Appendix
8. Learning Procedures
i. Introduction: The Virtues of Rescorla-Wagner
ii. Point Estimates of Causal Powers
iii. Adaptive Scores: The Bayesian Way
iv. Building on Patterns in the Data
v. Heuristics and Compromises
vi. Building on Sand
9. Representation and Rationality: The Case of Backwards Blocking
-
Representation
-
Backwards Blocking
-
Experiments
-
The Cheng Model
-
A Bayesian Calculation with Cheng Models
-
Bootstrapping to Backwards Blocking in Cheng Models
-
Alternative Bayes Net Representations
-
Conclusion
10. Cognitive Parts: from Freud to Farah
-
Parts, Beliefs and Habits: Classical Neuropsychology
-
The Connectionist Alternative
-
Freud
-
Farah
-
Issues
11. Inferring Cognitive Architecture from Individual Case
Studies
i. The Issues
-
Theories as Functional Diagrams and Graphs
-
Formalities
-
Discovery Problems and Success
-
An Illustration
-
Complications
-
Resource/PDP Models
12. Inferring Cognitive Architecture from Group Studies
-
Introduction
-
An Inexhaustive Review
-
Discovery Problems from Frequencies with Deterministic Input/Output Behavior
-
Discovery Problems with Indeterministic Input/Output Relations
-
Conclusion
13. The Explanatory Power of Lesioning Neural Nets
-
Introduction
-
Networks and Graphs
-
Feedforward Networks as Bayes Nets
-
Feedforward Networks without Hidden Nodes
-
Hidden Nodes
-
Recurrent Networks
-
Implications
Part 4: Social Psychology and Psychometrics
14. Social Statistics and Genuine Inquiry: The Case of The Bell Curve
-
Introduction
-
Varieties of Pseudo-Science
-
Inquiry and Discovery
-
The Bell Curve
-
Factor Analysis
-
Regression and Discovery
-
The Problems of Causal Inference
-
Some Empirical Cases
-
Projects and Attitudes
Preface
In a way, this book began thirty years ago. While I was at work on a methodological history of psychoanalysis, Leon Kamin gave me the manuscript of his book on intelligence testing, Science and IQ. That led me to read more on psychometrics and, increasingly, cognitive psychology. I came to the conclusion that most of the deficiencies in Freud’s science are deficiencies as well in contemporary theories of cognitive architecture, whether founded or psychometrics or on now more fashionable techniques in cognitive psychology, and so in a sense my history was unfair to Freud and his followers. I rewrote the manuscript as an historical parallelism with a methodological moral. While I thought its thesis was (and, sadly, still is) true, the result was thoroughly mean, and the unhappy manuscript sat in my cupboard with the dishes for several years, dragged out now and then for unsatisfactory repairs. In the meanwhile, my more productive work concerned the causal interpretation of Bayes nets and computational learning theory. In the 1980s, though conversations with several psychologists, especially Martha Farah, and with Jeff Bub, I began to think there were important but unnoticed connections between Bayes nets, computational learning theory, and contemporary cognitive psychology. In 1998 I succumbed to the usual vanity of philosophers and decided to publish a collection of my essays. No doubt as guilt payment for what must be thousands of hours of refereeing, MIT press agreed to publish the book after I assured them that every one of my relatives would buy at least one copy. One of the readers of the collection was pleased with the few essays on psychology I included, and thought a book that further developed their ideas would be valuable. As I began to prepare final revisions for publication, I concluded the reader was correct, and wrote what turned out to be a very different, and I hope more useful, book. Freud is still in the cupboard.
The methodological ideas in this book are indebted to Kevin Kelly, Dan Osherson, Judea Pearl, Thomas Richardson, Richard Scheines, Peter Spirtes, Gregory Cooper and Scott Weinstein. My application of these ideas to psychology is indebted to years of conversation with three psychologists, Patricia Cheng, Martha Farah and Alison Gopnik, and to more recent conversations with Susan Johnson, John Watson, and Joshua Tenenbaum. Chapter 3, in particular, emerged from joint work with Gopnik, and parts of Chapter 7 from joint work with Cheng, to whom I am especially grateful for her patient explanations and corrections of my many errors, mathematical and otherwise. Chapter 8 was prompted by discussions with Gopnik and Tenenbaum, and is really my side of a correspondence with Tenenbaum. In chapter 6, I applied some results of brilliant work on the Rescorla-Wagner model by my brilliant student, David Danks. Allen Hobson posed to me the problem addressed in Chapter 12. There is a pleasing somethng—irony? closure?—in the fact that aspects of Part II of this book are concerned with the Rescorla-Wagner model of learning, which was inspired by a suggestion of Kamin’s, and aspects of Part IV are concerned with inferences from intelligence tests, an issue to which Kamin first introduced me. So, although I have not met or talked with him in twenty five years, there is another debt.
Parts of Chapter 2 appeared in Synthese. A fragment of Chapter 3 appeared in “Bayes Nets as Psychological Models,” in a volume, Explanation and Cognition, edited by Frank Keil and Robert Wilson. Chapter 10 is an elaboration of an essay published in the British Journal for Philosophy of Science in 1994. Chapter 12 was written in collaboration with Thomas Richardson and Peter Spirtes and accepted for publication by Philosophy of Science some years ago, subject to elaborations we did not want to take the trouble to make, and which are unnecessary in the context of this volume. I thank Richardson and Sprites for permitting me to publish it here. Chapter 13 is an amalgam of two essays on the Bell Curve, one from Philosophy of Science and the other from a volume published by Springer-Verlag.
I owe an institutional debt to Carnegie Mellon University, to the National Aeronautics and Space Administration and to the NASA Ames Research Center, and to the National Science Foundation, which supported work on Part III of this book. I have a personal debt to Ken Ford, Ted Roush and Joseph Ramsey. Ford, who worked both at Ames and at the Insitute for Human and Machine Cognition at the University of West Florida, subpoenaed my assistance for a project at Ames on autonomous robotic mineral identification, which led to a grant from Ames that enabled Carnegie Mellon to free me from teaching duties for a year. Thanks to Roush, at Ames, and to the heroic efforts of Ramsey, who has been my programmer for the project, the work went very smoothly and fruitfully, and permitted me marginal time over three months to work on this book.
Share with your friends: |