Fact, Fiction, and Forecast
Fact, Fiction and Forecast
I. Nelson Goodman (wiki)
Henry Nelson Goodman (7 August 1906–25 November 1998) was an American philosopher, known for his work on counterfactuals, mereology, the problem of induction, irrealism, and aesthetics.
Goodman was born in Somerville, Massachusetts, the son of Sarah Elizabeth (née Woodbury) and Henry Lewis Goodman. He was of Jewish origins. He graduated from Harvard University, A.B., magna cum laude (1928). During the 1930s, he ran an art gallery in Boston, Massachusetts while studying for a Harvard Ph.D. in philosophy, which he completed in 1941. His experience as an art dealer helps explain his later turn towards aesthetics, where he became better known than in logic and analytic philosophy. During World War II, he served as a psychologist in the US Army.
He taught at the University of Pennsylvania, 1946–1964, where his students included Noam Chomsky, Sydney Morgenbesser, Stephen Stich, and Hilary Putnam. He was a research fellow at the Harvard Center for Cognitive Studies from 1962 to 1963 and was a professor at several universities from 1964 to 1967, before being appointed Professor of Philosophy at Harvard in 1968.
In 1967, at the Harvard Graduate School of Education, he was the founding director of Harvard Project Zero, a basic research project in artistic cognition and artistic education. He remained the director for four years and served as an informal adviser for many years thereafter.
Goodman died in Needham, Massachusetts.
II. Fact, Fiction, and Forecast
A) Editions (introductory notes)
The first edition (1954) consists of four lectures: the first (1946); the second, third, and fourth (1953)
The three rules formulated in the final chapter of the first edition are reduced to two in the second edition, and to one in the third edition (1973).
The third edition rewrites section 4 of chapter 4 and revises section 5 of chapter 4.
The fourth edition adds Putnam’s preface.
This book is translated from the fourth edition (Liu Huajie 2006)
B) Evaluation
Goodman perhaps understood more deeply than other philosophers that rejecting dualism fundamentally requires reconstructing the goals and means of philosophy / philosophers of science have long focused on the problem of induction and not on the problem of deduction; this asymmetric attitude is in fact problematic (translator’s afterword)
One consequence of Goodman’s discussion of the “new riddle of induction” is that it demotes a philosophical problem into a historical problem. / A second consequence is that it undermines a traditional view, namely that confirmation is a wholly logical relation between sentences (“An Historical Introduction to the Philosophy of Science,” chapter 13)
C) The problem of induction (chapter 3, sections 1–4)
- The old problem of induction
Hume held that “there is no necessary connection among matters of fact,” and that “custom is the great guide of human life.” Hume’s critics believed that his account applies only to the source of prediction, and not to the justification of prediction. Goodman, however, believes that Hume grasped the central issue, and that his answer is also valid, though not entirely satisfactory.
- How the old problem is dissolved
What can constitute a justification of induction?
Deduction is justified because it follows the general rules of deductive inference, and the general rules, in turn, are justified because they follow valid inferences. This is a valuable circle, whose point is that rules, like particular inferences, are justified by being placed in a process of mutual support. Likewise, if predictions follow valid inductive rules, then they are justified; if these rules precisely systematize accepted inductive practice, then they are valid. The problem of induction is not one of demonstrating validity, but of defining the difference between valid prediction and invalid prediction.
Class discussion: what counts as valid?
- The constructive task of confirmation theory
*Definitions can extend ordinary usage, or they can revise it.
The problem: define the relation holding between an arbitrary sentence S1 and another sentence S2 if and only if S1 can be properly said to confirm S2 to any degree whatsoever.
| Attempt | Problem |
| Confirmation includes the converse relation | All statements confirm all statements
Some statements are hypothetical consequences, but not all of their consequences confirm them; heterogeneous conjunction |
| A hypothesis is truly confirmed only by sentences that are instances of the hypothesis | The raven paradox
1) Unspoken and illicit reference to evidence not stated in the example 2) An error in the definition. It does not take all the evidence that has been stated into account. |
| We want, under certain restrictions, what is true in the narrow domain of evidence statements to be confirmed for the full domain of discourse | The new riddle of induction |
4) The new riddle of induction
Only lawlike sentences—whether true or false, and regardless of whether they are scientifically important—can have their instances confirmed; accidental sentences cannot.
| Example | Problem | Revision |
| Grue | Two incompatible hypotheses receive equal confirmation.
All predictions receive equal confirmation. (e.g. emerose) Anything confirms anything. |
1) Illicit information is used. (X)
2) Evidence works through other hypotheses. (lawlike) The circumstances related to one another must be precisely articulated. |
| Conductive | The problem of distinguishing lawlike hypotheses from accidental hypotheses reappears. | 1) Laws require complete universality (equivalent hypotheses, difficult to exclude)
2) Exclude locative predicates rather than quantitative ones. (X) |
//Wouldn’t it be better if we just discussed ordinary hypotheses? If we are seeking a theory, we should not evade these//
The new riddle of induction: how to distinguish lawlike or confirmable hypotheses from accidental or non-confirmable hypotheses.
- The generality problem concerning projection
The general problem of projection
III. The problem of counterfactuals, the problem of qualities, and the problem of possible constructs (chapters 1 and 2)
A) On philosophic conscience (chapter 2, section 1)
Goodman believes that philosophical problems require an acceptable account on an acceptable basis; we need to draw a boundary between what is already clear and what still needs clarification.
He opposes the verification theory of meaning, namely: “A sentence has meaning only if it can be empirically verified. Otherwise, it is either ‘analytic,’ a tautology; or, if neither empirical nor analytic, it is ‘metaphysical,’ that is, meaningless.” Goodman believes that the failure of this distinction leads to the libertine doctrine that “anything goes.”
Because there is no convenient, practicable criterion for “what counts as clear,” the thinker as an individual can only reflect on his own philosophic conscience. Yet each person’s philosophic conscience can yield only specific judgments, not general principles. (Choose his problematic research objects)
Below we will examine the problem of counterfactuals, the problem of qualities, and the problem of possible constructs
B) Counterfactual conditionals (chapter 1)
“If that piece of butter were heated to 150°F, it would not melt” and “If that piece of butter were heated to 150°F, it would melt” are both true because the antecedent is false, and this is something we cannot accept.
Class discussion: why can’t we accept this? The treatment in modal logic.
Counterfactual sentences, that is, sentences in which both the antecedent and the consequent are false. Since the antecedent is false, if we consider truth-functional composition, all counterfactual sentences are true.
Judging under what conditions a given counterfactual sentence holds, and under what conditions the corresponding contrary-to-fact conditional with contradictory result does not hold (the problem of counterfactuals), does not depend on the truth or falsity of the clauses, but on the expected relation of connection. This relation of connection may perhaps be explained by a general solution.
//Semifactuals are different.//
Goodman believes that the general problem of counterfactuals mainly has two aspects: the problem of relevant conditions and the problem of laws.
- The problem of relevant conditions
“The consequent (C) is derivable by law from the antecedent (A) and all true sentences (S)”
| Problem | Solution (mainly restricting S) | Notes |
| If one sentence in S is the negation of A, then A and S can derive anything. | Exclude from S sentences incompatible with A | ¬A∧A→C
S∧A |
| There are sentences in S that are logically compatible with A but otherwise incompatible | Counterfactuals cannot depend on empty laws | |
| If the above two conditions are satisfied (S is the set of all true sentences that are both logically compatible and non-logically compatible with A), then S∧A may also derive both C and ¬C, making it impossible to determine whether a counterfactual conditional is true or false. | Require that there be no S’, such that S’·A is self-consistent and can derive ¬C
|
|
| A possible objection to this solution: ordinarily there will be an S such that A·S is self-consistent and, by law, can derive ¬C | S must be simultaneously compatible with ¬C and with C. | ¬(A∧¬C)……Assume A and ¬C are incompatible
¬A∨C A→C A∧¬C……Assume A and ¬C are compatible Let S be ¬C S∧A ⇔ ¬C∧A→¬C
“Because ¬C is presumably true, S must necessarily be compatible with it” ¬(S∧¬C)……Assume S and ¬C are incompatible S→C……S is true, C is false, S→C is false; the assumption fails |
| S may include sentences that, although compatible with A, would not be true if A were true. Thus one can construct defective counterfactuals. | S and A are “jointly tenable” or “cotenable” | Additional note: the relative fixity of conditions is usually unclear |
Cotenability is itself a term defined for counterfactuals.
- The problem of laws
What is the distinction between the connecting principle and the relevant conditions? Is there some method for distinguishing laws from non-laws?
Causal facts / causal laws contingent fact instances
First approximation: a law is a true sentence used for prediction
Lawlike: sentences that satisfy the definition of a law regardless of whether they are true or false
Restated definition: a general sentence is lawlike if and only if it is acceptable before all its instances have been determined.
Dimo
Changed definition: a sentence is lawlike if its acceptability does not depend on the determination of any given instance.
Acceptability
“Some sentences do not require testing all positive outcomes; merely testing a single instance will lead us to accept that sentence and make predictions. Some sentences even positively determine several instances; this can lead us to accept the sentence with confidence and make predictions consistent with it.”
26 bullet billiards
What kind of predicates are projectible?
C）Counterfactuals, the Problem of Dispositions, and the Problem of Possible Concreta (Chapter 2, Section 234)
Counterfactual conditionals (counterfactual conditionals) eliminate disposition (disposition) terms without introducing any troublesome words such as “possible.”
- Counterfactual conditionals
Related condition problem / law problem?
Two reasons shift attention to dispositions: a) counterfactual conditionals make us focus more on the form of the sentence, which may be an obstacle; b) dispositional sentences may be simpler, because they only concern the problem of “internal states.”
- Disposition Q4
Manifest predicates and disposition predicates.
Problem of dispositions 1.0: the problem of explaining mysteriousness in manifest terms
Predicates merely refer to the things to which they apply; disposition predicates, like manifest predicates, apply only to actual things, and do not include non-actual things in their extension. The special feature of disposition predicates is that they apply to things that may occur rather than those that actually do occur.
Problem of dispositions 2.0: seeking a criterion for the correct assignment of things to disposition predicates formulated in actual-occurrence terms—that is, in manifest-predicate terms.
//First suggestion: a disposition predicate is only a summary description of certain specific aspects of an entity’s entire history.
Problem: assigning certain circumstances, inconsistent with the facts
The revision no longer confines us to things that actually occur; we can also discuss certain fictional things. Back to counterfactual conditionals.//
Bending/non-bending. Projecting this dichotomy onto broader categories or universal categories of things. The problem is how to define this projection using manifest predicates.
Kinds, essences.
A relation. Problem 2.1 of the problem of dispositions: characterize a relation such that if an initial manifest predicate “Q” stands in this relation to another manifest predicate “A,” or to the conjunction of manifest predicate “A,” then “A” may be equated with the dispositional counterpart of predicate “Q,” “Q’s”
Q5
//The problem of defining dispositional terms//
“Two points to keep in mind”: the formulation of the general problem / the recognition that dispositional as well as manifest predicates are labels used in classifying actual things.
- Possibility
This section discusses statements that explicitly refer to possible concreta beyond actual objects.
(Phenomenological) example: discussing the color problem of a nonexistent p-t concreta at time t and place p
| Attempt | Solution and its negation |
| Use a counterfactual conditional, assuming there is such a p-t | Discussed earlier; no good |
| There exists an actual concreta composed of p and t, a class {p, t} or an individual p+t | The parts lack a relation and do not constitute a “time-place” |
| “Possible ‘time-place’” as a disposition term projecting “time-place.”
eg. “‘Time-place’ p+t is scarlet” |
(Possible physical event) example: accidentable, may have had/might have had, mountainous London
Purpose: to translate statements about possibility into actual statements………The suggestion is that even discourse about possibles need not go beyond the actual world.
//Complex broad projection//
D）Transience (Chapter 2, Section 5)
Our problem: the projection problem, how to broaden predicates to a wider range
“The problem of induction”
IV. Projection Theory (Chapter 4)
A）A New Perspective on the Projection Problem
To solve the projection problem is to seek a general and precise method for saying which hypothesis is confirmed by a given body of evidence, or which projection can be validly drawn from given evidence.
In solving this problem, besides evidence and hypotheses, we can also make use of a great deal of information, and on the basis of actual projections define valid projection or projectibility.
What we are concerned with is not how the mind works, but the distinction it draws between valid projections and invalid projections.
B）Actual Projection
Actual projection involves the explicit and implicit formulation and adoption of hypotheses, and involves actual predictions of the results of further trials.
Whether a hypothesis is confirmed by a class of sentences depends on whether these sentences are truly accepted as evidential sentences, but our definition of the confirmation relation is independent of this consideration. The same goes for projectibility.
*positive instances or cases / negative instances or cases / undecided cases
*evidence class // projection class
*supported / violated / exhausted
“Only when a hypothesis, at the time in question, has certain undecided instances, certain positive instances, and no negative instances, does the adoption of a hypothesis constitute actual projection”
The double problem of defining projectibility: the unprojectible projected, and the projected unprojected.
C）Conflict Resolution
……
Elimination, though neither violated nor exhausted, of projected hypotheses that are non-lawlike.
The choice between green and grue: green is better entrenched than grue.
The entrenchment of a predicate arises from actual projection, but not merely from the actual projection of that predicate itself. Entrenchment is derived from the use of language.
The way to compare two conflicting projections is to see which is better entrenched; if one projection conflicts with a projection that is very well entrenched, then that projection ought to be eliminated.
Entrenchment is not the same as familiarity
Objection 1: legitimacy? A better entrenched predicate becomes a legitimate predicate.
Objection 2: truth? We do not know.
D）Presumptive Projectibility
A hypothesis is projectible if all conflicting hypothesis are overridden, unprojectible if overridden, and non-projectible if in conflict with another hypothesis and neither is overridden
Grund vs. grare: “When further evidence is needed to make a judgment between them, even the best entrenched hypothesis is non-projectible.”
All amorees are grays
Inherited entrenchment
Rule: a hypothesis is projectible if and only if it is supported, not violated, and not exhausted, and all hypotheses conflicting with it are defeated; it is non-projectible if and only if, together with a hypothesis conflicting with it, it is supported, not violated, not exhausted, and neither is defeated; it is unprojectible if and only if it is unsupported, violated, exhausted, or defeated.
E）Comparative Projectibility
There are three factors that lead to a low/high ultimate index: projectibility / the total number of supports / specificity
Positive overhypothesis: a hypothesis is a positive overhypothesis of a second hypothesis if the antecedent and consequent of the first hypothesis are the parent predicates of the antecedent and consequent of the second hypothesis, respectively.
It should be noted: 1) the number of positive instances of a hypothesis and its degree of projectibility are quite different factors; 2) projection is not transitive, but it can be strengthened/weakened; 3) “calculated” projectibility differs from the synthetic measure of projectibility.
“Crucial experiments”
Judgment of the actual situation
F）Summary and Conjecture
The source of inductive validity lies in our use of language; maintaining this consistency with already observed regularities is a function of our linguistic practice.
A proper interpretation of a counterfactual conditional should attend to its conflicts with other conditionals, and to the principles governing these conflicts.
A random problem
Not a mechanizable solution
Translated from the Chinese original with AI assistance. The original text is authoritative.
Leave a Reply