No label defined (Q5560255)
From MaRDI portal
!
This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:
scientific article; zbMATH DE number 3274713
| Language | Label | Description | Also known as |
|---|---|---|---|
| default for all languages | No label defined |
||
| English | No label defined |
scientific article; zbMATH DE number 3274713 |
Statements
1968
0 references
selected essays
0 references
0.7895699143409729
0 references
0.7513260841369629
0 references
The work contains an interesting selection of essays of partly mathematical, partly philosophical nature. The following are titles of the author's writings on inductive logic: (1) The relation between induction and probability. I, II. Mind 27, 389--404 (1918), 29 (1920). (2) Critical notice on J. M. Keynes' A treatise on probability. Mind 31 (1922). (3) Critical notice on W. E. Johnson's Logic. II. Mind 31 (1922). (4) Mr. Johnson on the logical foundations of science. I, II. Mind 33 (1924). (5) The philosophy of Francis Bacon (Cambridge, 1926). (6) The principle of problematic induction, Proc. Aristot. Soc. 28 (1927/8). (7) The principles of demonstrative induction, I, II, Mind 39, 302--317, 426--439 (1930). (8) Mechanical and teleological causation, Proc. Aristot. Soc., Suppl. Vol. 14 (1935). (9) Critical notice on R. von Mises, Wahrscheinlichkeit, Statistik und Wahrheit, Mind 46 (1937). (10) Hr. von Wright on the logic of induction. I--III. Mind 53, 1--24, 97--119, 193--214 (1944). (11) Critical notice on W. Kneale: Probability and Induction. Mind 59 (1950). All these essays with the exception of (5) and (10) are reprinted in the present book, which, in addition contains a paper by G. H. von Wright ``Broad on induction and probability'' and a last paper by Broad ``Reply to my critics''. It is not easy for a mathematician to do justice to a philosopher but the reviewer shall try it although it is difficult to match the author's beautiful style and his tolerance. As to the author, ``all induction, however advanced and complicated ultimately rests on `induction by simple enumeration' and on the use of the `hypothetical method'\,''. The first is of the type: ``All observed \(A\)'s have been \(B\). Therefore all \(A\)'s whatever will be \(B\).'' We gladly agree with the author that this conclusion is wrong. The author goes on stating that most inductive logicians think that the fallacy can be avoided by the introduction of an additional premise called the Uniformity of Nature or Law of Causation. He shows, however, that this will not work and he proposes the formulation: ``On the observed data, it is probable (or highly probable) that all \(A\)'s are \(B\).'' That means: Inductive conclusions are to be in terms of probability. In order to justify this he proceeds to consider the drawing of counters out of a bag. The bag contains \(n\) counters of two colors white and black in unknown proportions. \(m\) counters are drawn without replacement, all white. By means of the ``principle of indifference'' he concludes that ``all mutually exclusive alternatives'' are equally probable and obtains Laplace's rule of succession, which states that the probability for the next counter to be drawn being white equals \((m+1)/(m + 2)\) and the probability that all \(n\) are white equals \((m+1)/(n+1)\). This works if the\( n+1\) possible proportions are assumed equally probable. In (6), seven years later, he returns to the same problems and he as well as Keynes, whom he quotes, severely criticize Laplace for this assumption, speaking of the ``false Laplacean assumption'' which must ``be rejected''. He does not seem to realize that Laplace makes a mathematical assumption and draws mathematical conclusions in a correct way. The author proves then that this bag problem, even part of the question of uniform prior distribution, cannot be applied to nature, to the color of craws, e.g., since many of the assumptions valid for the counters do not work for the craws. This again we gladly believe. He considers more problems on inverse probability which are mathematically far from new. The really interesting and valuable limit theorems due to Laplace, to von Mises and also to the reviewer (inference on so-called rare events) which might even be of help in the ``natural'' cases are apparently not known to the author. All he wants -- in agreement with Keynes -- is to find rules for correct application of the indifference principle and although by keeping to these rules some obvious mistakes might be avoided the ``principle'' just is too narrow. (The difficulty of an adequate ``probability'' has been overcome by two different approaches: Kolmogorov's and von Mises'.) Much more interesting and original seems his contribution to the ``hypothetical method''. Let \(h\) be a hypothesis. One proves by ordinary deductive reasoning that \(h\) implies certain propositions \(c_1, c_2,\ldots, c_n\). Let \(h\vert f\) be the probability of \(h\) on the basis of evidence \(f\) collected before starting the new investigation. It is then easily seen \(h\vert c_1c_2\cdots c_nf=(h\vert f)/c_1c_2\cdots c_n\vert f\). The initial probability \(h\vert f\), which must be positive, is increased through the division by \(c_1c_2\cdots c_n\). If, which is certainly greater than \(h\vert f\) and which equals the product of \(n\) factors, each less than 1 (unless some \(c_i\) implies another). The numerator is the more increased the greater \(n\) (number of verified consequences which are logically independent on each other) and it is the more increased the smaller the factors like \(c_1\vert f\) or \(c_2\vert f\) etc. A hypothesis is indeed greatly strengthened if it leads to some surprising consequence which is found to be true. A similar theorem has been proved by Keynes in a more complicated way and later than the author. This ``Fundamental Theorem on Confirmation'', if new, is, the reviewer thinks a very worthwhile achievement. It happens repeatedly that the author gives at different places different opinions on the same matter. An example: In the review (9) the author admits that, as Mises explains, the use of Bernoulli's theorem as a ``bridge'' between Laplace's and Mises' definition of probability is fallacious. But this same Bridge, used by Keynes, is not criticized. Now a point regarding (9). The author is understandably concerned whether Mises' two postulates (existence of frequency limit; randomness) are compatible. He finds (p. 195) the logical objections against the notion of ``collective'' plausible and Mises' answers ``not very convincing''. We add that in 1936, A. Wald (and partly before him A. Copeland) have mathematically proved that, with a careful formulation the two postulates (and therefore the concept ``collective'') are consistent. It must be admitted that the knowledge of these sophisticated mathematical considerations was not the author's business. However, next the author objects that neither the existence of a frequency limit nor that of insensitivity to ``place selections'' can be seen empirically!! In a mathematical journal this problem needs no answer. The author also says that von Mises ``claims'' to have constructed so-called Bernoullian sequences. These are however published in Math. Ann. 1933. We have said that it is not always easy for mathematicians to do justice to philosophers. May the reviewer add that the philosophers should be more careful to do justice to mathematicians. On p. 67 the author says, agreeing with Keynes, that Laplace's rule of succession, in case the counters are replaced after each trial holds only if the number of drawings tends to infinity, an assumption which becomes unnecessary for drawing without replacement. Actually the state is as follows: Following Laplace -- whom they however severely reprimand -- Broad and Keynes assume in both cases equal probability for the possible proportions. (See the author (6).) If this is assumed then Laplace's rule holds in both cases and without passage to infinity. If, however, the number of drawings tends to infinity then the formula holds even without the above assumption of uniformity, i. e., independent on the ``prior distribution''. Whether Laplace has proved this last statement explicitly the reviewer could not say. It holds certainly true and can be proved in a similar way as other limit theorems on inverse probability.
0 references