Inductive Logic and Science Author(s): Rudolf Carnap Reviewed work(s): Source: Proceedings of the American Academy of Arts

and Sciences, Vol. 80, No. 3, Contributions to the Analysis and Synthesis of Knowledge 3 (Mar., 1953), pp. 189-197 Published by: American Academy of Arts & Sciences Stable URL: http://www.jstor.org/stable/20023651 . Accessed: 01/03/2012 16:31
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

American Academy of Arts & Sciences is collaborating with JSTOR to digitize, preserve and extend access to Proceedings of the American Academy of Arts and Sciences.

http://www.jstor.org

INDUCTIVE LOGIC AND SCIENCE

Eudolf Carnap1

of the usefulness of inductive The question logic for science of the very possibility and even the question of inductive logic are today still debated. The conception differs here explained from that of many other scholars. that inductive First I wish to emphasize logic does not propose new ways to explicate of thinking, but merely old ways. It im certain forms of reasoning tries to make which explicit or instinctively have always been applied both in every plicitly in This is analogous to the situation life science. at and day of deductive the beginning did not invent logic. Aristotle that had gone on as long ago as there deductive reasoning; " : What was human language. If somebody had said to Aristotle have done well good is your new theory to us? We enough should we change our ways of thinking without it. Why and he might have answered: "I do accept your new inventionV', not propose new ways of thinking, I merely want to help you to do consciously and hence with and safety greater clarity from pitfalls what you have always done. I merely want to
replace common sense by exact rules."

It is the same with inductive Inductive logic. mean I likewise as old as human here language.
reasoning clusion all goes forms beyond of the reasoning content or of inference the premises,

is reasoning inductive by
the and con there

where

fore cannot be stated with certainty. if a physicist states a new law or a theory Thus, for example, as a system of laws on the basis of the experimental results an inductive he makes he has found, inference. So does a an unknown scientist who assumes single fact on the basis of for example, the meteorologist known facts; who predicts the
weather for tomorrow, the physicist who assumes a certain

of the velocities of gas molecules which he cannot a to historian who tries observe, directly explain a reported act a certain of Abraham Lincoln by hypothetically assuming distribution

i read at the Paper Soc. of the Am. Phys. 1951. 27 October

on Philosophy of Symposium and the Philosophy of Science

session joint Science, Association, Chicago,

190
motivation
an estimate

carnap

which
for the

is not
unknown

reported,
value

or a statistician
of a parameter

who makes
in a popula

tion on the basis of an observed sample from the population. common ways of to explicate intends Since inductive logic merely the question of its usefulness inductive leads back reasoning, to the general question: Is it desirable that procedures which or instinctively, are generally though only intuitively applied, are brought into the clear daylight, and systematized analyzed answer in the form of exact rules ?Whoever gives an affirmative of to this general will the importance acknowledge question of explicating inductive that is, the special problem reasoning, a system of inductive of constructing logic with rules as exact as those of the older, well-established system of deductive logic. as a solution for this Whether any particular system proposed and fruitful is workable is, of course, another question. problem As I see it, the fundamental of inductive concept logic is All is inductive probability. reasoning probability reasoning. is not unambiguous. I refer the word 'probability' However, in one particular here to probability sense, the logical sense, call inductive This concept must which we might probability. sense. from probability in the statistical be clearly distinguished is practically The distinction and theoretically important
fundamental.

is a certain quantitative char Statistical probability physical of physical acteristic systems. Like any other physical magni In tude it is to be established empirically, by observations. are of a statistical nature. this case the observations They is in counting Statistical consist frequencies. probability connected with but it is not very closely frequency, obviously we say that for a given die When just the same as frequency. an ace is 0.158, of throwing then this state the probability ment refers to a physical of the die and thus is characteristic from statements not fundamentally about its mass, different etc. Imagine a fictitious phys electric conductivity, temperature, icist who, like the Laplacean superman, knows the present micro state of the die in terms of the distribution of the particles all the relevant laws. This and of the fields and, in addition, find not calculations, could, by purely mathematical physicist its of the the temperature etc., die, conductivity, present only an ace if thrown under of its yielding but also the probability is actually not known, Since the micro-state specified conditions. about the probability, the question arises how to test a statement in the case of it. The answer how to confirm or disconfirm

inductive

logic

and

science

191

from that in the case different is not fundamentally probability or other physical of temperature The statement magnitudes. is to be tested by making which arrangements experimental connected with the magnitude lead to observable phenomena To in question, whose value observable. itself is not directly we fre test the probability the determine relative statement, of aces in a sufficiently quency long series of throws of the is itself not the probability; die. This frequency it is rather a consequence state of the die, a consequence of the probability is observable and therefore may serve for us as a symp which tom for the probability of the state, just as the expansion is not itself the temperature column in the thermometer mercury
but an observable consequence of the temperature state and there

fore a suitable means of testing a statement about the tempera ture. It is sometimes said that the statistical concept of prob no finite a peculiar since obviously involves difficulty, ability to determine series of throws is sufficient the probability with and certainty. This is indeed true, but the absolute precision same holds for all physical magnitudes. is likewise no There for determining the with procedure temperature possible and certainty. to the question: The answer absolute precision the series of throws with the "How long then shall we make ?" is the same as die in order to determine the probability the answer to the question: "How fine a thermometer should we use to measure In both cases the answer the temperature?" on the one hand, on the time and money available depends, on on other the the desired of and, degree hand, precision. on the theoretical or practical More it depends specifically, to be expected from higher precision. The finer the advantages thermometer and the longer the series of throws, the higher case is there a per the precision which is achieved. In neither fect procedure. The concept of statistical may be probability introduced either by an explicit definition in terms of a limit as done by Mises and Reichenbach, or by an axiom system with as done by the majority rules of application of contemporary In either case, the concept is logically statisticians. legitimate and practically useful for work in statistics and in all branches of science which apply statistical methods. Thus I do not agree with of the inductive those representatives of prob concept like Keynes and Jeffreys, who the statistical ability, reject On the other hand, I do not agree with Mises, Eeichen concept. and the statisticians, who reject the inductive bach, concept. are important Both concepts of probability for scientific work,

192

CARNAP

science itself, the other in each in its own field, the one within with inductive logic, which gives rules for certain operations con of science. The statistical of the language the statements is certain today generally cept recognized. problems Although a defense of connected with it are still under investigation, and usefulness is no longer necessary. The status its legitimacy of the inductive is still debated. Therefore, concept, however, to its it is still and to defend necessary today right of existence show its usefulness. states a relation between A statement of inductive probability a hypothesis and a given body of evidence, e.g., results of actual or possible The asserted probability value means observations. or supported to which the hypothesis is confirmed the degree on to notice that a statement It is important by the evidence. or degree to is relative inductive of confirmation probability is that the statement This does not merely mean the evidence. from observations. That is the case for based on or derived For example, statement. the statement "The every scientific rain tomorrow is 1/5" that it will is incomplete probability to such and such an evidence", unless we add "with respect For of observations. the validity e.g., certain meteorological it does not matter whether referred the evidence the statement to in the statement is true and whether it is known to the speaker. of any inference, whether To be sure, in the practical application or inductive, are usually the premises known. But deductive
that is not necessary. They may be unknown or they may even

If this is so, what can be the basis of the to be false. statement the of itself, as distinguished validity probability or the evidence? of the hypothesis It can from the validity not be of an empirical nature. All relevant empirical obviously or assumption in the evidence statement. is contained knowledge the truth of the In our example, we can empirically reexamine be known
evidence concerning past meteorological events. If we wait un

test the truth of the hypothesis til tomorrow, we can empirically But in neither way can we test the that it will rain tomorrow. statement truth of the probability shall see tomor itself. We
row either rain or not-rain, we may observe a rain of short or of

a rain of high or low intensity, but we shall not see long duration, a rain of probability Some critics of inductive 1/5. logic have to and from it the fact drawn conclusion since this that, pointed statement is not empirically the inductive testable, probability was that Their mistake it must be scientifically meaningless. as a statement the factual synthetic they regarded probability

inductive

logic

and

science

193

statement. The statement is, however, of a purely logical nature. Hence of empirical there is no need and no possibility testing. A statement of inductive probability is in one respect similar to a statement in deductive between the hy logic: the relation and the evidence which it asserts is a logical relation, pothesis or incompati of deducibility similar to the deductive relations a asserts than If the statement weaker those. bility, though value close to 1, then the probability relation hereby probability is very close to the relation of deducibility: the hy expressed is from the evidence but not quite. On nearly deducible pothesis the other hand, if the stated probability value is near to 0, then the probability relation is close to the deductive relation of in : the hypothesis is nearly with the compatibility incompatible evidence but not quite. For any intermediate value probability remote relation is more the probability from the deductive
relations which are, so to speak, the extreme cases. Thus induc

means in a sense partial tive probability It is a deducibility. as a de as can relation inasmuch it be logical established, just ductive relation, as soon as the two statements of hypothesis and evidence are given, by merely in this applying logical analysis, case the rules of inductive the use of observa logic, without tions. Although the statement expresses only a logical relation, it has nevertheless It draws boundaries to reason significance. able conduct. For example, if the probability of rain tomorrow to the evidence is 1/5 with available to an observer, respect then it would not be reasonable for him to bet on rain tomorrow at odds higher than 1:4. on inductive probability If we recognize that statements have a purely are in a position we then to clear logical character, for two hundred up a question which has been debated years: the problem of the so-called principle reason or of insufficient As I see it, the beginning of the de principle of indifference. of in the classical inductive velopment logic was made theory of probability and Laplace. by men like Bernoulli, Bayes, Many points of the classical theory, and among them also fundamental for more than a hundred years and points, have been criticized our in I think that this criticism is correct century. especially to a large extent. I agree with the critics that today it is im to go back to the classical I do not But possible conception. with those who that the out the total is agree say only way of the classical The classical of rejection conception. principle are known which "If no reasons indifference states: would favor one of several possible events, then the events are to be

194

CARNAP

The usual objection taken as equally probable." this against on ignorance; if you do is that it puts a premium principle not know anything about the alternatives, then the principle a certain if you allows you to make statement about them; no statement is know certain then that longer permis things, looks rather absurd. from ignorance sible. To derive a statement to apply this procedure to a And indeed be absurd it would statement. But to of equipr ob ability factual the statement of indifference leads is, like all statements the principle which on inductive not a factual but a logical statement. probability, If the knowledge of the observer does not favor any of the as evidence to then with this knowledge events, respect possible The statement they are equally probable. equal prob assigning in this case does not assert anything about the facts, abilities the between but merely about the logical relations something of these each the that evidence and namely, hypotheses; given are logically alike. These relations are obviously relations alike to structure if the evidence has a symmetrical with respect asserts events. The statement of equiprobability the possible than the symmetry. nothing more I believe that the basic idea of the old these reasons For of indifference is valid. On the other hand, there principle can be no doubt that many of the applications of this principle, and that some in the earlier period, were invalid especially even outright I absurd. But drawn were of the conclusions the classical pioneers believe that the aim which envisaged was con the classical valid. Our task is not to abandon entirely an con to exact of of construct but concept degree ception, of inductive prob the classical firmation conception explicating ability on a more cautious and more solid foundation. let us look at the function of inductive Now logic in the on the one hand, observa field of science. A scientist makes, or of results of experiments. These tions of natural phenomena he entertains the other his evidence. On constitute hand, facts not yet observed or even unobserv concerning hypotheses able. The hypothesis may concern a single fact or it may be a to do such of the form "If we were conditional prediction then such and such events would and such things, happen", or it may have a general form, e.g., a statement about the value or a general law stating the relations of a material constant, in terms of mathematical between various physical magnitudes form or a statis The law may have a deterministic functions.
tical form, stating e.g., proportions, averages, or other statistical

INDUCTIVE

logic

and

science

195

The pur of distributions of certain magnitudes. parameters and predict new pose of the law is to explain known phenomena ones. The task of inductive logic is not to find a law for the of given phenomena. This task cannot be solved explanation or by fixed rules; it is rather by any mechanical procedure and the good luck solved through the intuition, the inspiration, of the scientist. The function of inductive logic begins after a is offered for examination. Its task is to measure hypothesis the support which the given evidence supplies for the tentatively In particular, the task will assumed often be to hypothesis. one which determine several the among competing hypotheses is most confirmed The com evidence. strongly by the given concern the possible results peting hypotheses may, for example, to be made, of an experiment the possible causes of an observed or a business outcomes of investment. Or they event, possible are mutually laws which each of may be various incompatible, as an explanation which might be regarded of a given set of results concerning new phenomena observational not explained
so far.

an objection Sometimes is raised against the idea of a sys tem of inductive exact rules for the determination logic with of the degree of confirmation because of the fact that a scientist who chooses one among a number of considered is hypotheses influenced in this choice also by many non-rational factors and to hand over the task of this not be willing that he would
choice to a machine or to have himself, so-to-speak, transformed

into a machine which merely fixed rules. Now it is applies true that many non-rational factors affect the scientist's choice, and I believe that this will always be the case. The influence of some of these factors may be undesirable, for instance a bias in favor of a hypothesis maintained or, previously publicly in the case of a hypothesis in social science, a bias caused by or political moral But there are also non-rational preferences. effect is important factors whose and fruitful; for example, the influence of the "scientific or hunch". instinct Inductive factors of this kind. Its func logic does not intend to eliminate a clearer picture to give to the scientist tion is merely of the to what degree the various hypoth situation by demonstrating eses considered are confirmed This by the evidence. logical influence by inductive picture supplied logic will (or should) the scientist, but it does not uniquely determine his decision of the choice of a hypothesis. He will be helped in this decision in the same way a tourist is helped by a good map. If he uses

196

CARNAP

inductive still remains his; it will, however, logic, the decision be an enlightened decision rather than a more or less blind one. In addition to judging the status of hypotheses, inductive rules of estimation. There logic has also the task of supplying
is much discussion and controversy among statisticians concern

of particular and the methods of estimation ing the validity choice of a suitable method in a given problem of estimation I believe that, if the basis of inductive situation. logic is con structed the degree of by laying down rules for calculating to define a general then it is possible estimate confirmation, in terms of degree function of confirmation, to all applicable in the language kinds of magnitudes in question. expressible I propose The definition takes as the estimate which of the on a the basis of magnitude given body of evidence the weighted mean of the magnitude, of of the possible values the weight each value being its degree of confirmation with respect to the This is the same as the expectation value of the given evidence. in the term lexpectation value' magnitude (if we understand sense based on inductive contrast in the inductive probability, sense based on statistical to its statistical To obtain probability). a general method of estimation would be of great importance not only from a theoretical point of view but also for the prob in a rational way. Sup lem of determining decisions practical
pose tion, choose a man has to make a decision in This of alternative a e.g., concerning one among investments. a number given means economic that actions he possible situa has to to

him in the situation. For each possible the action he considers outcomes various in terms of money If he is possible gained. for each possible the degree of confirmation able to determine outcome he calculate in the case of the considered may action, with its degree of con the sum of these gains, each multiplied of the gain in the case of This will be his estimate firmation. In the same way, he may calculate the the action considered. actions. of the gain for each of the possible estimate Then, if he is a rational man, he will choose that one among the possible actions for which the estimated (A gain has its greatest value. more exact procedure would consider, not the gain in terms of of satis but the utility of this gain, i.e., the measure money, Thus inductive faction derived by the person from the gain.) of rational for the determination logic serves as an instrument
decisions.

As mentioned earlier, the development the classical theory of probability. with

of inductive However,

logic began its systema

INDUCTIVE

LOGIC

AND

SCIENCE

197

as a branch of modern tization origin, be logic is of recent thirty years ago. I have Keynes ginning with John Maynard a set of rules of inductive constructed logic for a simple language of things to is restricted which descriptions qualitative system, without the use of measurable (like temperature, magnitudes of frequencies. statements electric current, etc.), but including of the degree of con the calculation These rules make possible and any body of evidence express for any hypothesis firmation of the estimate ible in that language system and the calculation on the basis of any given evidence. The further of a frequency more for of inductive language comprehensive logic development re of science as a whole systems and finally for the language a task for the future. mains

A theory in my book Logical of inductive logic is systematically developed Foundations 1950. The underlying of of Probability, Chicago, conception in The Nature and Application inductive is explained of In probability ductive is a reprint of six non-technical sections Logic, 1951, which Chicago, from the book mentioned.