Models, Truth, and Realism模式、真理、實在論

1、models, truth, and realismtaylor, barry, university of melbourneabstract: this book mounts an argument against one of the fundamental tenets of much contemporary philosophy, the idea that we can make sense of reality as existing objectively, independently of our capacities to come to know it. part o

2、ne argues that traditional realism can be explicated as a doctrine about truth that truth is objective, that is, public, bivalent, and epistemically independent. part two argues that a form of hilary putnams model-theoretic argument demonstrates that no such notion of truth can be founded on the ide

3、a of correspondence, as explained in model-theoretic terms. part three argues that non-correspondence accounts of truth-truth as superassertibility or idealized rational acceptability, formal conceptions of truth, and tarskian truth also fail to meet the criteria for objectivity. along the way, it a

4、lso dismisses the claims of the latter-day views of putnam, and of similar views articulated by john mcdowell, to constitute a new, less traditional, form of realism. the coda bolsters some of the considerations advanced in part three in evaluating formal conceptions of truth, by assessing and rejec

5、ting the claims of robert brandom to have combined such an account of truth with a satisfactory account of semantic structure. the book concludes that there is no defensible notion of truth that preserves the theses of traditional realism, nor any extant position sufficiently true to the ideals of t

6、hat doctrine to inherit its title. so the only question remaining is which form of antirealism to adopt.contentsoverview: the argument of this book 1part i the explication of realism 111 realism and objective truth 132 realism explicated 32part ii model theory and correspondence 483 putnams model-th

7、eoretic arguments 494 changing the rules 855 the status of natural properties 101part iii realism without correspondence? 1256 taking the hierarchy seriously 1257 formal theories of truth and putnams common-sense realism 1338 tarskian truth and the views of john mcdowell 148preface this is a short b

8、ook, but it is the product of more years of thought on its topics than i care to enumerate. it covers a good deal of ground, but is written in rather a compressed style; so, on the advice of the press, i have included an overview, which i hope will be of use to a reader in following the twists and t

9、urns of the books argument.chapter 5 in large part reproduces a paper i first published in mind (on natural properties in metaphysics, mind, 102 (1993), 81100); its provenance is hereby acknowledged. the argument of chapter 3 also has its origins in a former paper ( “just more theory”: a manuvre in

10、putnams model-theoretic argument for antirealism, australasian journal of philosophy, 69 (1991), 15266), which i now find barely intelligible myself. this version is entirely rewritten, considerably enlarged, and, hopefully, also considerably improved.this work was assisted by a research grant from

11、the australian research council in 20001, which is gratefully acknowledged. thanks are due for comments from members of various seminars at the universities of melbourne and adelaide, monash and la trobe universities, and the australian national university, where versions of various parts of this wo

12、rk were presented; and for comments from two anonymous readers for the press. finally, special thanks are due to my colleagues graham priest and greg restall, both of whom read and commented on an earlier draft of the book in its entirety; and extra special thanks to the second of these, who, a glut

13、ton for punishment, did the same thing again for a revised version. despite their efforts of all the above, i am afraid i must assume full responsibility for the defects which remain.b. m. t.melbourne, april 2006 overview: the argument of this book this book falls into three parts, along with a coda

14、. part one argues that traditional realism can be explicated as a doctrine about truththat truth is objective, that is, public, bivalent, and epistemically independent. part two argues that a form of hilary putnams model-theoretic argument demonstrates that no such notion of truth can be founded on

15、the idea of correspondence, as explained in model-theoretic terms (more traditional accounts of correspondence having been already disposed of in part one). part three argues that non-correspondence accounts of truthtruth as superassertibility or idealized rational acceptability, formal conceptions

16、of truth, tarskian truthalso fail to meet the criteria for objectivity; along the way, it also dismisses the claims of the latter-day views of putnam, and of similar views articulated by john mcdowell, to constitute a new, less traditional, form of realism. the coda bolsters some of the consideratio

17、ns advanced in part three in evaluating formal conceptions of truth, by assessing and rejecting the claims of robert brandom to have combined such an account of truth with a satisfactory account of semantic structure. the book concludes that there is no defensible notion of truth that preserves the

18、theses of traditional realism, nor any extant position sufficiently true to the ideals of that doctrine to inherit its title. so the only question remaining is which form of antirealism to adopt.part onethe strategy of this part is to show, by extrapolation and generalization, how traditional realis

19、m is associated with objective truth, characterized as above.seeking a neutral and uncontroversial starting-point, chapter 1 begins by examining and adapting a characterization of realism offered by michael devitt, settling on an initial formulation (section 1.1) of realism about objects of kind k a

20、s the doctrine that objects of kind k exist objectively, and explaining objective existence in terms of intersubjective warrant. the latter notion is structured by assuming a framework of possible epistemic standpoints, modelled on the points of view of leibnizean monads, and each based upon a possi

21、ble course of experience of the world; these standpoints certify beliefs through the application of idealized epistemology to the experiences they embody. then objects of kind k are said to exist objectively iff they exist, and the belief that they exist can be certified from at least two such stand

22、points; and the doctrine that they do exist objectively in this sense is object realism about the objects of kind k.but (section 1.2) object realism is an inadequate framework for the discussion of some debates between realists and their opponents. to do these justice, we need to invoke the apparatu

23、s of situations and facts (obtaining situations). then fact realism concerning a range of situations is the doctrine that these situations are objective facts, i.e., that they obtain and can be certified to obtain from at least two epistemic standpoints. object realism is subsumed under fact realism

24、 as a special case.however (section 1.3), the argument discussed in the literature under various colourful soubriquetsthe great fact argument, the slingshotreveals that the apparatus of situations is too fragile to bear any explanatory weight. but the developing analysis can be freed from reliance o

25、n this apparatus by semantic ascent. applying this technique, fact realism is supplanted by semantic realism, where semantic realism concerning a set of sentences of a language l is the doctrine that these sentences are true-in-l, and can be certified to be true-in-l from at least two epistemic stan

26、dpoints.section 1.4 simplifies this definition by replacing the truth requirement, so that the whole definition is cast in terms of certification. it appeals to the convergence thesis, according to which good epistemology homes in on the objective truth as information increases. to put this to use,

27、it extends the framework of epistemic standpoints, enabling any finite union of standpoints to count as a standpoint, so that the standpoints form a hierarchy of certification with more information available to higher level standpoints. then convergence means that a true sentence will eventually be

28、stably certified as true by some standpoint s (i.e., certified as true by s and its union with any other standpoint). a notion of corroborated certification is introduced to play, within the context of the extended framework of standpoints, the role previously filled by the requirement of multiple c

29、ertification. then the new definition of semantic realism about a set of sentences of l is that it holds all such sentences are objectively true in the sense that they have a stable, corroborated certification as true-in-l from some standpoint s.chapter 2 sets out to extract from this account more o

30、f the properties of objective truth. to do so (section 2.1), it inquires into the principles governing certification. principles governing the logical connectives are borrowed from crispin wright, and tested for adequacy against intuition, particularly on the question of the distributivity of object

31、ive truth across the connectives. this investigation suggests, first, the introduction of further principles governing second-order certification; and, further, the extension of the hierarchy of certification to a further, infinite level to be inhabited by the total view, comprising the union of all

32、 finite standpoints. this permits further simplifying the definition of objective truth, so that it becomes just certification from the total view. pure realism concerning a set of sentences of l becomes the doctrine that all members of the set are objectively true-in-l, and global realism about l t

33、he doctrine that pure realism holds for some maximal set of sentences of l.section 2.2 now extracts from this final account of objective truth that objective truth is public, bivalent, and epistemically independent (the last specifically in the sense that a theory ideal by the standards of human epi

34、stemology might be objectively false). finally, section 2.3 suggests that the framework of standpoints and its hierarchy of certification, exploited to obtain these insights, can be treated merely as a useful device rather than a metaphysical reality, in much the way possible worlds are treated by m

35、any as no more than a useful fiction for investigating modality. thus we can explicate objective truth as truth with the hallmark traits, without commitment to the literal truth of the framework which has helped us to identify them. the question for realism is whether there is some defensible notion

36、 of truth possessing these traits, and applicable to a language apt for framing serious theory.part twothe time-honoured way for realists to found objective truth is by appeal to correspondence. the traditional account of correspondence, however, was discredited back in section 1.3, along with the f

37、acts and situations it invokes as the second terms of that relation. but model theory, in the spirit of the old account, also explains truth in terms of relations between language and extralinguistic structures, and can be regarded as a conceptually hygienic, precisely formulated, embodiment of the

38、old intuitions. putnam has, however, advanced a case for thinking model-theoretic truth unsuited as an account of objective truth as we have explicated that notion, a case which this part evaluates.chapter 3 sets out what we take to be the relevant core of putnams case. section 3.1 extracts three ar

39、guments from putnams writings: the arguments from cardinality, completeness, and permutation. of these, section 3.2 argues, only the second is of direct relevance. appealing to the completeness theorem, it shows that the ideal theory, provided it can be cast in a first-order language, must have a mo

40、del m, whence its theses must in some sense be true (since they must be true-on-m), thus threatening epistemic independence. the other two arguments bear less directly on the realism issue as explicated in part one. nevertheless, all three arguments in their disparate ways raise the same issue, name

41、ly, how to distinguish intended models from the others. if this distinction can be made secure, then, in particular, the argument from completeness can be defused, provided the completeness-guaranteed model m turns out to be unintended, so that truth-on-m is not a serious brand of truth.at issue, ac

42、cordingly, is the possibility of framing constraints which will sort out the intended models. the most popular approach is to cast such constraints as reference constraints, which seek to restrict the models which satisfy them to those which assign to linguistic terms entities which are, intuitively

43、, their correct referents. one favourite way to devise such constraints is to exploit a reductionist account of reference, and section 3.3 examines attempts to frame constraints based on causal and psycho-behavioural reductions of reference. section 3.4 investigates the translational reference const

44、raint (trc), a constraint on reference which does not rely on a reduction of reference but makes essential use of translation (from object language to metalanguage) to sort out the models which get reference right. trc in turn suggests a stronger constraint on intendedness, the translational truth c

45、onstraint (ttc), which goes beyond merely constraining reference, being equivalent to trc combined with a constraint on the domain on quantification, the translational domain constraint (tdc). ttc, it is argued, constitutes the gold standard of constraints on intendedness. the claims made in this se

46、ction, however, require foundation in a theory of translation, sufficient to sustain the assumptions it makes about that controversial and opaque notion. this foundation is supplied in section 3.5, whose general tenor is davidsonian, its key notion being that of a hermeneutic theory, i.e., a davidso

47、nian theory of interpretation cast into model-theoretic terms.with ttc now identified as the most fundamental constraint on intendedness, it remains to see if it will suffice to rule out as unintended all the models of ideal theory whose existence the completeness theorem guarantees. the issue is ex

48、amined in section 3.6. a notorious argument of putnams, the just more theory manuvre, argues that no constraint on intendedness whatever can do this, because a statement of the constraint can be added to the ideal theory, and completeness guarantees that this enlargement of the ideal theory will its

49、elf have a model. following david lewis, many find this manuvre involves a confusion between satisfying a constraint, and making true a statement which claims to express it.upon investigation, it emerges that lewiss discussion oversimplifies what is involved, because of the metalinguistic nature of

50、constraints on intendedness in general, and of ttc in particular. nevertheless, taking these complications into account, the locus of the fallacy lewis claims to detect can be identified. still, in the special case in which we are interested, where it is models of ideal theory which are in question,

51、 no fallacy is involved, because of the restricted nature of the metalanguages of the language of ideal theory which we can meaningfully suppose to be available. thus the argument from completeness remains unimpugned by ttc, and other constraints on intendedness which it subsumes.chapter 4 investiga

52、tes other ways the argument from completeness might be attacked. section 4.1 investigates challenging that arguments assumption that first-order resources suffice for framing the ideal theory. it concedes that there may indeed be reasons for thinking first-order resources must be extended in some wa

53、y if ideal theory is to be accommodated, but argues that there arealmostno grounds for thinking the extension will mean the unprovability of completeness, as it must if the conclusion of the argument is to be avoided. the hedge in the almost is to acknowledge that there is a case, albeit an underwhe

54、lming one, for casting ideal theory as a second-order language, for which completeness is often held to fail. but, it is argued, the judgement that it fails depends on dismissing as illegitimate and unintended the henkin model theory for second-order languages; and there is no way of justifying that

55、 dismissal which does not beg the question.section 4.2 looks at a suggestion by g. h. merrill, endorsed by david lewis, that the model theory used to frame the definition of objective truth should take a nonstandard form, with the referents of predicates restricted to elite sets, the extensions of n

56、atural properties marking inbuilt cleavages in nature independent of human psychology and convention. since the proposed model theory is nonstandard, the suggestion is, completeness will no longer hold, and the problems it engenders will vanish. a problem arises about the formulation of this suggest

57、ion. lewis suggests it can be recast, not as a proposal to adopt nonstandard model theory but as the requirement that standard models satisfy an eliteness constraint, which he counts as a constraint on intendedness. for us, this raises the question of the relation of the proposed eliteness constrain

58、t to the fundamental constraint on intendedness, ttc, and its vulnerability or otherwise to the just more theory manuvre. the upshot of the discussion is that merrills proposal can, indeed, be formulated using an eliteness constraint in combination with the fundamental constraint on intendedness ttc

59、, and that the eliteness constraint itself should be regarded not as a constraint on intendedness but as an additional, metaphysical constraint; and, further, that it is at least arguable that it is not itself subject to the reasoning of the just more theory manuvre, since it makes demands on the resources of the metalanguage beyond those made by constraints on intendedness, thus altering the relations between objec