ABSTRACT
The object of this chapter is to suggest that, although prominent contributions on the F9LMJ=9F<MK=>MDF=KKG>=;GFGE=LJA;K@9N=E9<=KA?FAd;9FLJ=>=J=F;=LG&=QF=KcNA=OK GFL@AKLGHA;K=== ? +9LAFCAF #=F<JQ ALJ=E9AFKL@=;9K=L@9L&=QF=Kc EGJ=b>MF<9E=FL9Dc;JALA;AKEKG>=;GFGE=LJA;K9J=AFL@=E9AFF=?D=;L=< /@MKO@ADKL KM;@;GFLJA:MLAGFK@9N=L=F<=<LG;GF;=FLJ9L=GF&=QF=Kc;GEE=FLKGF=;GFGE=LJA;KL@9L <=9D OAL@ bL=;@FA;9D AKKM=Kc KM;@ 9K AFEG<=JF L=JEAFGDG?Q HJG:D=EK G> GEALL=< variable bias, simultaneous equation bias and so on), I wish to argue that the main thrust G>&=QF=Kc;JALA;AKEK9J=G>9EGJ=H@ADGKGH@A;9DGJE=L@G<GDG?A;9DF9LMJ=<=JANAF?>JGE the theories and arguments which he developed in his Treatise on Probability. $L AK 9DKG OGJL@ FGLAF? L@9L KLM<A=K O@A;@ <G AF >9;L =P9EAF= &=QF=Kc A<=9K 9K
presented in the Treatise on Probability tend not to relate them to his later views on =;GFGE=LJA;K $F<==< 9 J=;=FL J=D9LAN=DQ=PL=FKAN= ;GFLJA:MLAGF :Q *cGFF=DD O@A;@ <G=KLJ9;=L@=J=D9LAGFK@AH:=LO==F&=QF=KcH@ADGKGH@A;9D9F<GL@=JOJALAF?K9<EALKL@9L 9b<=L9AD=<<AK;MKKAGFG>&=QF=KL@=GJQG>AF<M;LAGF6O@A;@$K@9DD9J?M=AKF=;=KK9JQAF GJ<=JLGMF<=JKL9F<@AK;GEE=FLKGF=;GFGE=LJA;K7@9K:==FGEALL=<c 9F< *cGFF=DDKM??=KLKL@9L9b>MJL@=JKLM<Q`&=QF=K+@ADGKGH@Q9F< ;GFGE=LJA;KaOGMD< :=9>MJL@=JOGJL@O@AD==FL=JHJAK=c /@AK;@9HL=JL@=J=>GJ=;9F:=AFL=JHJ=L=<=AL@=J9K9F9LL=EHLLGdDDH9JLG>9?9HGJ
?AN=F J=;=FL 9;;GMFLKG>&=QF=Kc NA=OKGF =;GFGE=LJA;K 9F 9LL=EHL LG J=<J=KK 9F AE:9D9F;= G> =EH@9KAK /G L@AK =F< &=QF=Kc9;;GMFL G> HJG:9:ADALQ 9F< AF<M;LAGF AK summarised below and some implications of this account are illustrated in the important ;GFL=PLG>bHJ=<A;LAGFc9F<bHJ=<A;LAN=9;;MJ9;Qc !GDDGOAF?GF>JGEL@AK9F9KK=KKE=FLAK E9<=G>L@=9KKMEHLAGFKL@9L9;;GJ<AF?LG&=QF=K9J=>MF<9E=FL9DA>AF<M;LAN=E=L@G<K 9J=LG:==EHDGQ=<D=?ALAE9L=DQ /@=J=D=N9F;=G>L@=K=AKKM=KLG&=QF=Kc;JALA;AKEKG> =;GFGE=LJA;KAK9F9DQK=<F=PL9F<dF9DDQKGE=;GF;DMKAGFK9J=<J9OF
Keynes and the ‘Treatise on Probability’
&=QF=KcL@=GJQG>HJG:9:ADALQJ=HJ=K=FLK9E9BGJ<=N=DGHE=FLOAL@AFL@=DG?A;9DLJ9<ALAGF of probability. In this account, even if it is not possible to argue demonstrably from any one proposition to another one, there still exists a logical relation between the two propositions. This is the probability relation. If a conclusion a is related to a premise h with probability p then the probability relation-sometimes called the argument-is written a/h=p. The probability of the conclusion a is always relative to some evidence h. The discovery of new evidence h1 does not =FL9ADL@9LL@=HJ=NAGMKb9J?ME=FLcO9K wrong, it just gives rise to a new argument a/h k.
