Ideas from 'Foundations without Foundationalism' by Stewart Shapiro [1991], by Theme Structure

[found in 'Foundations without Foundationalism' by Shapiro,Stewart [OUP 1991,0-19-825029-0]].

green numbers give full details    |     back to texts     |     unexpand these ideas


3. Truth / F. Semantic Truth / 1. Tarski's Truth / b. Satisfaction and truth
Satisfaction is 'truth in a model', which is a model of 'truth'
                        Full Idea: In a sense, satisfaction is the notion of 'truth in a model', and (as Hodes 1984 elegantly puts it) 'truth in a model' is a model of 'truth'.
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], 1.1)
                        A reaction: So we can say that Tarski doesn't offer a definition of truth itself, but replaces it with a 'model' of truth.
4. Formal Logic / A. Syllogistic Logic / 1. Aristotelian Logic
Aristotelian logic is complete
                        Full Idea: Aristotelian logic is complete.
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], 2.5)
                        A reaction: [He cites Corcoran 1972]
4. Formal Logic / F. Set Theory ST / 3. Types of Set / a. Types of set
A set is 'transitive' if contains every member of each of its members
                        Full Idea: If, for every b∈d, a∈b entails that a∈d, the d is said to be 'transitive'. In other words, d is transitive if it contains every member of each of its members.
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], 4.2)
                        A reaction: The alternative would be that the members of the set are subsets, but the members of those subsets are not themselves members of the higher-level set.
4. Formal Logic / F. Set Theory ST / 4. Axioms for Sets / j. Axiom of Choice IX
Choice is essential for proving downward Löwenheim-Skolem
                        Full Idea: The axiom of choice is essential for proving the downward Löwenheim-Skolem Theorem.
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], 4.1)
4. Formal Logic / F. Set Theory ST / 5. Conceptions of Set / a. Sets as existing
Are sets part of logic, or part of mathematics?
                        Full Idea: Is there a notion of set in the jurisdiction of logic, or does it belong to mathematics proper?
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], Pref)
                        A reaction: It immediately strikes me that they might be neither. I don't see that relations between well-defined groups of things must involve number, and I don't see that mapping the relations must intrinsically involve logical consequence or inference.
4. Formal Logic / F. Set Theory ST / 5. Conceptions of Set / e. Iterative sets
Russell's paradox shows that there are classes which are not iterative sets
                        Full Idea: The argument behind Russell's paradox shows that in set theory there are logical sets (i.e. classes) that are not iterative sets.
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], 1.3)
                        A reaction: In his preface, Shapiro expresses doubts about the idea of a 'logical set'. Hence the theorists like the iterative hierarchy because it is well-founded and under control, not because it is comprehensive in scope. See all of pp.19-20.
It is central to the iterative conception that membership is well-founded, with no infinite descending chains
                        Full Idea: In set theory it is central to the iterative conception that the membership relation is well-founded, ...which means there are no infinite descending chains from any relation.
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], 5.1.4)
Iterative sets are not Boolean; the complement of an iterative set is not an iterative sets
                        Full Idea: Iterative sets do not exhibit a Boolean structure, because the complement of an iterative set is not itself an iterative set.
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], 7.1)
4. Formal Logic / F. Set Theory ST / 6. Ordering in Sets
'Well-ordering' of a set is an irreflexive, transitive, and binary relation with a least element
                        Full Idea: A 'well-ordering' of a set X is an irreflexive, transitive, and binary relation on X in which every non-empty subset of X has a least element.
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], 5.1.3)
                        A reaction: So there is a beginning, an ongoing sequence, and no retracing of steps.
5. Theory of Logic / A. Overview of Logic / 1. Overview of Logic
There is no 'correct' logic for natural languages
                        Full Idea: There is no question of finding the 'correct' or 'true' logic underlying a part of natural language.
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], Pref)
                        A reaction: One needs the context of Shapiro's defence of second-order logic to see his reasons for this. Call me romantic, but I retain faith that there is one true logic. The Kennedy Assassination problem - can't see the truth because drowning in evidence.
Logic is the ideal for learning new propositions on the basis of others
                        Full Idea: A logic can be seen as the ideal of what may be called 'relative justification', the process of coming to know some propositions on the basis of others.
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], 2.3.1)
                        A reaction: This seems to be the modern idea of logic, as opposed to identification of a set of 'logical truths' from which eternal necessities (such as mathematics) can be derived. 'Know' implies that they are true - which conclusions may not be.
5. Theory of Logic / A. Overview of Logic / 2. History of Logic
Bernays (1918) formulated and proved the completeness of propositional logic
                        Full Idea: Bernays (1918) formulated and proved the completeness of propositional logic, the first precise solution as part of the Hilbert programme.
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], 7.2.1)
Can one develop set theory first, then derive numbers, or are numbers more basic?
                        Full Idea: In 1910 Weyl observed that set theory seemed to presuppose natural numbers, and he regarded numbers as more fundamental than sets, as did Fraenkel. Dedekind had developed set theory independently, and used it to formulate numbers.
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], 7.2.2)
Skolem and Gödel championed first-order, and Zermelo, Hilbert, and Bernays championed higher-order
                        Full Idea: Skolem and Gödel were the main proponents of first-order languages. The higher-order language 'opposition' was championed by Zermelo, Hilbert, and Bernays.
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], 7.2)
5. Theory of Logic / A. Overview of Logic / 5. First-Order Logic
First-order logic was an afterthought in the development of modern logic
                        Full Idea: Almost all the systems developed in the first part of the twentieth century are higher-order; first-order logic was an afterthought.
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], 7.1)
The 'triumph' of first-order logic may be related to logicism and the Hilbert programme, which failed
                        Full Idea: The 'triumph' of first-order logic may be related to the remnants of failed foundationalist programmes early this century - logicism and the Hilbert programme.
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], Pref)
                        A reaction: Being complete must also be one of its attractions, and Quine seems to like it because of its minimal ontological commitment.
Maybe compactness, semantic effectiveness, and the Löwenheim-Skolem properties are desirable
                        Full Idea: Tharp (1975) suggested that compactness, semantic effectiveness, and the Löwenheim-Skolem properties are consequences of features one would want a logic to have.
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], 6.5)
                        A reaction: I like this proposal, though Shapiro is strongly against. We keep extending our logic so that we can prove new things, but why should we assume that we can prove everything? That's just what Gödel suggests that we should give up on.
The notion of finitude is actually built into first-order languages
                        Full Idea: The notion of finitude is explicitly 'built in' to the systems of first-order languages in one way or another.
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], 9.1)
                        A reaction: Personally I am inclined to think that they are none the worse for that. No one had even thought of all these lovely infinities before 1870, and now we are supposed to change our logic (our actual logic!) to accommodate them. Cf quantum logic.
5. Theory of Logic / A. Overview of Logic / 7. Second-Order Logic
Second-order logic is better than set theory, since it only adds relations and operations, and nothing else
                        Full Idea: Shapiro preferred second-order logic to set theory because second-order logic refers only to the relations and operations in a domain, and not to the other things that set-theory brings with it - other domains, higher-order relations, and so forth.
                        From: report of Stewart Shapiro (Foundations without Foundationalism [1991]) by Shaughan Lavine - Understanding the Infinite VII.4
Broad standard semantics, or Henkin semantics with a subclass, or many-sorted first-order semantics?
                        Full Idea: Three systems of semantics for second-order languages: 'standard semantics' (variables cover all relations and functions), 'Henkin semantics' (relations and functions are a subclass) and 'first-order semantics' (many-sorted domains for variable-types).
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], Pref)
                        A reaction: [my summary]
Henkin semantics has separate variables ranging over the relations and over the functions
                        Full Idea: In 'Henkin' semantics, in a given model the relation variables range over a fixed collection of relations D on the domain, and the function variables range over a collection of functions F on the domain.
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], 3.3)
In standard semantics for second-order logic, a single domain fixes the ranges for the variables
                        Full Idea: In the standard semantics of second-order logic, by fixing a domain one thereby fixes the range of both the first-order variables and the second-order variables. There is no further 'interpreting' to be done.
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], 3.3)
                        A reaction: This contrasts with 'Henkin' semantics (Idea 13650), or first-order semantics, which involve more than one domain of quantification.
Completeness, Compactness and Löwenheim-Skolem fail in second-order standard semantics
                        Full Idea: The counterparts of Completeness, Compactness and the Löwenheim-Skolem theorems all fail for second-order languages with standard semantics, but hold for Henkin or first-order semantics. Hence such logics are much like first-order logic.
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], 4.1)
                        A reaction: Shapiro votes for the standard semantics, because he wants the greater expressive power, especially for the characterization of infinite structures.
5. Theory of Logic / B. Logical Consequence / 4. Semantic Consequence |=
Semantic consequence is ineffective in second-order logic
                        Full Idea: It follows from Gödel's incompleteness theorem that the semantic consequence relation of second-order logic is not effective. For example, the set of logical truths of any second-order logic is not recursively enumerable. It is not even arithmetic.
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], Pref)
                        A reaction: I don't fully understand this, but it sounds rather major, and a good reason to avoid second-order logic (despite Shapiro's proselytising). See Peter Smith on 'effectively enumerable'.
If a logic is incomplete, its semantic consequence relation is not effective
                        Full Idea: Second-order logic is inherently incomplete, so its semantic consequence relation is not effective.
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], 1.2.1)
5. Theory of Logic / E. Structures of Logic / 1. Logical Form
Finding the logical form of a sentence is difficult, and there are no criteria of correctness
                        Full Idea: It is sometimes difficult to find a formula that is a suitable counterpart of a particular sentence of natural language, and there is no acclaimed criterion for what counts as a good, or even acceptable, 'translation'.
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], 1.1)
5. Theory of Logic / G. Quantification / 4. Substitutional Quantification
We might reduce ontology by using truth of sentences and terms, instead of using objects satisfying models
                        Full Idea: The main role of substitutional semantics is to reduce ontology. As an alternative to model-theoretic semantics for formal languages, the idea is to replace the 'satisfaction' relation of formulas (by objects) with the 'truth' of sentences (using terms).
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], 9.1.4)
                        A reaction: I find this very appealing, and Ruth Barcan Marcus is the person to look at. My intuition is that logic should have no ontology at all, as it is just about how inference works, not about how things are. Shapiro offers a compromise.
5. Theory of Logic / I. Semantics of Logic / 4. Satisfaction
'Satisfaction' is a function from models, assignments, and formulas to {true,false}
                        Full Idea: The 'satisfaction' relation may be thought of as a function from models, assignments, and formulas to the truth values {true,false}.
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], 1.1)
                        A reaction: This at least makes clear that satisfaction is not the same as truth. Now you have to understand how Tarski can define truth in terms of satisfaction.
5. Theory of Logic / J. Model Theory in Logic / 1. Logical Models
Semantics for models uses set-theory
                        Full Idea: Typically, model-theoretic semantics is formulated in set theory.
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], 2.5.1)
5. Theory of Logic / J. Model Theory in Logic / 2. Isomorphisms
Categoricity can't be reached in a first-order language
                        Full Idea: Categoricity cannot be attained in a first-order language.
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], 7.3)
An axiomatization is 'categorical' if its models are isomorphic, so there is really only one interpretation
                        Full Idea: An axiomatization is 'categorical' if all its models are isomorphic to one another; ..hence it has 'essentially only one' interpretation [Veblen 1904].
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], 1.2.1)
5. Theory of Logic / J. Model Theory in Logic / 3. Löwenheim-Skolem Theorems
The Löwenheim-Skolem theorems show an explosion of infinite models, so 1st-order is useless for infinity
                        Full Idea: The Löwenheim-Skolem theorems mean that no first-order theory with an infinite model is categorical. If Γ has an infinite model, then it has a model of every infinite cardinality. So first-order languages cannot characterize infinite structures.
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], 4.1)
                        A reaction: So much of the debate about different logics hinges on characterizing 'infinite structures' - whatever they are! Shapiro is a leading structuralist in mathematics, so he wants second-order logic to help with his project.
Downward Löwenheim-Skolem: each satisfiable countable set always has countable models
                        Full Idea: A language has the Downward Löwenheim-Skolem property if each satisfiable countable set of sentences has a model whose domain is at most countable.
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], 6.5)
                        A reaction: This means you can't employ an infinite model to represent a fact about a countable set.
Upward Löwenheim-Skolem: each infinite model has infinite models of all sizes
                        Full Idea: A language has the Upward Löwenheim-Skolem property if for each set of sentences whose model has an infinite domain, then it has a model at least as big as each infinite cardinal.
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], 6.5)
                        A reaction: This means you can't have a countable model to represent a fact about infinite sets.
Substitutional semantics only has countably many terms, so Upward Löwenheim-Skolem trivially fails
                        Full Idea: The Upward Löwenheim-Skolem theorem fails (trivially) with substitutional semantics. If there are only countably many terms of the language, then there are no uncountable substitution models.
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], 9.1.4)
                        A reaction: Better and better. See Idea 13674. Why postulate more objects than you can possibly name? I'm even suspicious of all real numbers, because you can't properly define them in finite terms. Shapiro objects that the uncountable can't be characterized.
5. Theory of Logic / K. Features of Logics / 3. Soundness
'Weakly sound' if every theorem is a logical truth; 'sound' if every deduction is a semantic consequence
                        Full Idea: A logic is 'weakly sound' if every theorem is a logical truth, and 'strongly sound', or simply 'sound', if every deduction from Γ is a semantic consequence of Γ. Soundness indicates that the deductive system is faithful to the semantics.
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], 1.1)
                        A reaction: Similarly, 'weakly complete' is when every logical truth is a theorem.
5. Theory of Logic / K. Features of Logics / 4. Completeness
We can live well without completeness in logic
                        Full Idea: We can live without completeness in logic, and live well.
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], Pref)
                        A reaction: This is the kind of heady suggestion that American philosophers love to make. Sounds OK to me, though. Our ability to draw good inferences should be expected to outrun our ability to actually prove them. Completeness is for wimps.
5. Theory of Logic / K. Features of Logics / 6. Compactness
Non-compactness is a strength of second-order logic, enabling characterisation of infinite structures
                        Full Idea: It is sometimes said that non-compactness is a defect of second-order logic, but it is a consequence of a crucial strength - its ability to give categorical characterisations of infinite structures.
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], Pref)
                        A reaction: The dispute between fans of first- and second-order may hinge on their attitude to the infinite. I note that Skolem, who was not keen on the infinite, stuck to first-order. Should we launch a new Skolemite Crusade?
Compactness is derived from soundness and completeness
                        Full Idea: Compactness is a corollary of soundness and completeness. If Γ is not satisfiable, then, by completeness, Γ is not consistent. But the deductions contain only finite premises. So a finite subset shows the inconsistency.
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], 4.1)
                        A reaction: [this is abbreviated, but a proof of compactness] Since all worthwhile logics are sound, this effectively means that completeness entails compactness.
5. Theory of Logic / K. Features of Logics / 9. Expressibility
A language is 'semantically effective' if its logical truths are recursively enumerable
                        Full Idea: A logical language is 'semantically effective' if the collection of logically true sentences is a recursively enumerable set of strings.
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], 6.5)
6. Mathematics / A. Nature of Mathematics / 3. Nature of Numbers / b. Types of number
Complex numbers can be defined as reals, which are defined as rationals, then integers, then naturals
                        Full Idea: 'Definitions' of integers as pairs of naturals, rationals as pairs of integers, reals as Cauchy sequences of rationals, and complex numbers as pairs of reals are reductive foundations of various fields.
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], 2.1)
                        A reaction: On p.30 (bottom) Shapiro objects that in the process of reduction the numbers acquire properties they didn't have before.
6. Mathematics / A. Nature of Mathematics / 3. Nature of Numbers / d. Natural numbers
Only higher-order languages can specify that 0,1,2,... are all the natural numbers that there are
                        Full Idea: The main problem of characterizing the natural numbers is to state, somehow, that 0,1,2,.... are all the numbers that there are. We have seen that this can be accomplished with a higher-order language, but not in a first-order language.
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], 9.1.4)
6. Mathematics / A. Nature of Mathematics / 3. Nature of Numbers / e. Ordinal numbers
Natural numbers are the finite ordinals, and integers are equivalence classes of pairs of finite ordinals
                        Full Idea: By convention, the natural numbers are the finite ordinals, the integers are certain equivalence classes of pairs of finite ordinals, etc.
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], 9.3)
6. Mathematics / A. Nature of Mathematics / 5. The Infinite / g. Continuum Hypothesis
The 'continuum' is the cardinality of the powerset of a denumerably infinite set
                        Full Idea: The 'continuum' is the cardinality of the powerset of a denumerably infinite set.
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], 5.1.2)
6. Mathematics / B. Foundations for Mathematics / 4. Axioms for Number / d. Peano arithmetic
First-order arithmetic can't even represent basic number theory
                        Full Idea: Few theorists consider first-order arithmetic to be an adequate representation of even basic number theory.
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], 5 n28)
                        A reaction: This will be because of Idea 13656. Even 'basic' number theory will include all sorts of vast infinities, and that seems to be where the trouble is.
6. Mathematics / B. Foundations for Mathematics / 6. Mathematics as Set Theory / a. Mathematics is set theory
Some sets of natural numbers are definable in set-theory but not in arithmetic
                        Full Idea: There are sets of natural numbers definable in set-theory but not in arithmetic.
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], 5.3.3)
6. Mathematics / C. Sources of Mathematics / 6. Logicism / c. Neo-logicism
Logicism is distinctive in seeking a universal language, and denying that logic is a series of abstractions
                        Full Idea: It is claimed that aiming at a universal language for all contexts, and the thesis that logic does not involve a process of abstraction, separates the logicists from algebraists and mathematicians, and also from modern model theory.
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], 7.1)
                        A reaction: I am intuitively drawn to the idea that logic is essentially the result of a series of abstractions, so this gives me a further reason not to be a logicist. Shapiro cites Goldfarb 1979 and van Heijenoort 1967. Logicists reduce abstraction to logic.
6. Mathematics / C. Sources of Mathematics / 6. Logicism / d. Logicism critique
Mathematics and logic have no border, and logic must involve mathematics and its ontology
                        Full Idea: I extend Quinean holism to logic itself; there is no sharp border between mathematics and logic, especially the logic of mathematics. One cannot expect to do logic without incorporating some mathematics and accepting at least some of its ontology.
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], Pref)
                        A reaction: I have strong sales resistance to this proposal. Mathematics may have hijacked logic and warped it for its own evil purposes, but if logic is just the study of inferences then it must be more general than to apply specifically to mathematics.
6. Mathematics / C. Sources of Mathematics / 10. Constructivism / d. Predicativism
Some reject formal properties if they are not defined, or defined impredicatively
                        Full Idea: Some authors (Poincaré and Russell, for example) were disposed to reject properties that are not definable, or are definable only impredicatively.
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], 7.1)
                        A reaction: I take Quine to be the culmination of this line of thought, with his general rejection of 'attributes' in logic and in metaphysics.
8. Modes of Existence / B. Properties / 10. Properties as Predicates
Properties are often seen as intensional; equiangular and equilateral are different, despite identity of objects
                        Full Idea: Properties are often taken to be intensional; equiangular and equilateral are thought to be different properties of triangles, even though any triangle is equilateral if and only if it is equiangular.
                        From: Stewart Shapiro (Foundations without Foundationalism [1991], 1.3)
                        A reaction: Many logicians seem to want to treat properties as sets of objects (red being just the set of red things), but this looks like a desperate desire to say everything in first-order logic, where only objects are available to quantify over.