Future Directions and Conclusions
What has been accomplished in this book? First and foremost, I have proposed a theory of learning causal relationships that integrates two different sources of information: the experiences the learner observes and the knowledge the learner possesses when the observations are made. Furthermore, the theory claims that analytical learning techniques that make use of prior knowledge are to be preferred to empirical learning techniques when both are applicable. Section 7.7 reviews the psychological evidence that supports the claim that people exhibit this same preference. In Chapters 6 and 3, I indicate the computational reasons for this preference. In addition, I have argued that people possess a general theory of causality to assist in learning causal relationships, and demonstrated how a machine learning system can benefit from this source of knowledge.