1: Interesting but incomplete
Schum and Starale have put together some very interesting commentary and analysis of questions about the role played by evidence related issues in the history of probability and decision science,as well as current evaluations concerning evidence in recent work done in fuzzy logic,possibility theory,Bayesian conditionalization,networks,etc.
The most interesting chapter is chapter 5 .Its major flaw is a very incomplete overview of Keynes's contributions to questions about the role played by evidence in probability and decision theory in chapter 26 of the TP CONCERNING HIS INDEX,W, TO MEASURE THE COMPLETENESS OF THE EVIDENCE as well as Keynes's decision rule,c,the conventional coefficient of risk and weight;the other major omission is the failure to provide any discussion whatsoever of the important contribution made by D.Ellsberg in his famous 1961 article on ambiguity in decision making in the Quarterly Journal of Economics.The book's positive aspects are a very wide ranging coverage of the grading of probabilities,the standard numerical,tree diagram Pascalian probabilities,L Jonathan Cohen's Keynes like Baconian probabilities in section 5.5,Dempster(Shafer)belief functions,precision versus imprecision in probabilistic assessment and elicitation, fuzzy set theory, and fuzzy logic theory.
The authors are certainly correct when they conclude that Keynes was the first to explicitly differentiate between the weight and the probability of an outcome(see p.256).Carnap's 1950 belief that Peirce was the first overlooks the fact that the standard error of the estimate is a probabilisticly related term.OF course,Keynes discussed the fact that hypotheses with high evidential weight also usually would have minimal standard errors of the estimate,meaning that the standard error could sometimes be used as an approximation of the weight in Science related areas where there is a great deal of replication and duplication of work where the basic phenomenon studied consists of data which is stable,uniform,homogeneous and repetitive throughout time
On the other hand,the authors commit a stunning error in their claim(see p.314) that Keynes used a single Pascalian probability to grade the credibility-related characteristics of an evidence source.They completely overlook the interval valued nature of Keynesian probabilities presented by Keynes in chapters 15 and 17 of his 1921 A Treatise on Probability(TP).THEY CONSIDER ONLY CHAPTER 16 IN ISOLATION FROM THE REST OF THE BOOK.The failure to integrate a discussion of Keynes's chapter 26 extension of the chapter 6 accounts of the weight of the argument(evidence)is most likely the reason why this error was committed .This error means that the authors have no real understanding of the fundamental foundations underlying Keynes's general,logical approach to probability.The Pascalian calculus has a role to play but it is subservient to the Baconian approach.
|
