Quinean Naturalized Epistemology


Conceptual Project (What is the project that Carnap attempted and failed to carry through?)

      Carnap’s failure shows the impossibility of translating statements about bodies into statements about experience.  Recall Quine’s holism.



Doctrinal Project (Where did Hume leave us?  In skepticism about bodies, about causes, about the future, etc.)









The Conceptual and Doctrinal Projects were part of the Proof Paradigm:  Epistemology as Top-Down.

(1) The Cartesian Project (described by Kornblith on pp. 322-3):  (a) foundational beliefs and (b) set of epistemic principles of inference.

(2) Rational Reconstruction:  Top-Down justification of beliefs about bodies (Hume's problem) or of scientific theories (Carnap's problem).


Quine's Naturalized Epistemology is a Response to the Perceived Failures of the Conceptual and Doctrinal Projects of Traditional Epistemologists—the failure of epistemology as “first philosophy”.  (What does this mean?)






Quine’s Proposal:  Epistemology becomes a chapter of empirical psychology, where the goal is "to understand the link between between observation and science."(290)

[Need not assume that psychology is limited by Quine's behaviorist tendencies (e.g., his equation of evidence with “the stimulation of sensory receptors”.] 


“Epistemology still goes on, though in a new setting and a clarified status. Epistemology, or something like it, simply falls into place as a chapter of psychology and hence of natural science”(292).










Three Potential Problems:


(1) The Circularity Problem: "the old threat of circularity"(293).

Why does Quine think that evolution can help to clarify induction?


(2) The Normativity Problem:  Justification and Knowledge are normative (they concern what I ought to believe); psychology is purely descriptive.  Can there be epistemology without anything normative?


(3) BonJour’s Intellectual Suicide Objection. Quine is giving up a priori justification.  How can he do that without committing intellectual suicide?


Kornblith's Naturalized Epistemology


Three Epistemological Projects: 


(1) Account of HUMAN knowledge and justification.


(2) Show the extent of our knowledge.


(3) Provide epistemic advice (how to get more true beliefs and fewer false ones).













Cartesian Foundationalism is a "Good Argument" Account.  It is a fruitless research project.  It provides no good answers to any of the three questions that motivated it.


Essential components of naturalized epistemology:

(1) Rejection of a priori knowledge.

(2) Realization that skeptical questions arise within science. 


At least much of epistemology becomes an empirical discipline.

(Why "at least much"? (p. 326))








Kornblith’s rejection of universality (compare Zagzebski):  The topic is human knowledge, not an analysis of the concept of knowledge:  “Knowledge is a natural phenomenon”(326).


What is human knowledge?  “Reliably produced belief”(327).


Human knowledge is produced by belief producing mechanisms that are well-adapted to this world.


Examples:  (1) Visual illusion of motion (328).

            (2) Generalizing from small samples (328).







Kornblith thinks it is a mistake to search for principles of reasoning which would work in any possible world.  Our concern is with reliability in this world.


"The good-making features of the psychological mechanisms need not be such that they would tend to produce true beliefs in every possible world; rather, they only be well-adapted to this world" (325).


According to Kornblith, the reliability of our cognitive processes is deeply contingent (329).


What about the reasoning that leads to the discovery of visual illusions, cognitive biases, and other mistake-correcting reasoning?


Three Objections


I.  What is Kornblith's answer to the Circularity Problem?


II.  What is Kornblith's answer to the Normativitity Problem?


III.  BonJour’s Intellectual Suicide Objection:  Is naturalized epistemology self-referentially inconsistent?


Kornblith’s reply:  "Naturalists believe that human beings are so provided by nature that they are inclined to make certain kind of inferences which are in fact reliable, long before they have evidence that those inferences are reliable. .  . . BonJour is simply taking for granted certain constraints on good reasoning which the naturalist rejects”(332).


This response is developed by Nozick.

To evaluate the Kornblith/Nozick response, we need to distinguish universality from a prioricity: 


(1) Universality.  Are there universal concepts of knowledge and justification?

Are there universal principles of good reasoning?


(2) A Prioricity.  If so, how are they knowable?  Could they be discovered, rather than known a priori?














Nozick’s Question About Reasons:  What is the nature of the connection between a reason (r) and what it is a reason for (h)?


Three ways of answering the question:


(1) On the a priori approach, there is an objective logical relation of support between r and h that we directly apprehend.


(2) On the simple factual approach, there is an objective factual relation between r and h such that when r is true, h is likely to be true.







(3) On Nozick's sophisticated factual approach, the following addition is made to the simple factual approach:  Evolution has selected for beings to whom it seems self-evident that when r is true h is likely to be true.  (Evolution selects for seeming self-evidence [i.e., the apparent a priori].)


“Reason tells us about reality because reality shapes reason.”(112)

Example of Euclidean geometry.

      Selection for approximate truth.

      Serviceability rather than truth.


“To explain why such principles seem self-evident to us, one need not invoke their necessity.”(111)


Necessary truths are not necessary to explain why we believe in them!




Kant's Copernican Revolution and Nozick's reversal of it.


What is Nozick's response to the Problem of Induction?





















"The strength and depth of our intuitions about certain statements cannot be used as powerful evidence for their necessity if those statements are of a kind that, were they contingent facts, would have led to selection favoring strong intuitions of their self-evidence."(111)














Brandom’s Linguistic Rationalism

A Parochial Communitarian Internalist (Pragmatic) Epistemology


      Semantics as use in a rule-governed game.  (What is the game?)


Declarative sentences are normatively evaluated in two ways:

      (1) (subjective/deontological appraisal): Whether the speaker followed the rules of the game so as not to be blameworthy for producing the assertion.  Involves inferential consequences and inferential antecedents.

      (2) (objective appraisal):  Whether the assertion is correct in that things really are as it says they are.  (Truth = objective representational correctness) 






The Language Game of Epistemology


The game of giving and asking for reasons has two kinds of rules:

      (1) rules of consequential commitment (the committive dimension) [these correspond to truth conditions]

      (2) rules of entitlement (the critical dimension) [these correspond to the requirements for justification]


      (a) Committive.  Commitment-preserving inference generalizes deductive inference;

      (b) Permissive.  Entitlement-preserving inference generalizes inductive inference;

      (c) Incompatibility Entailments.  Generalizes modal (counter-factual supporting) inference.  (e.g., If there were no oxygen in this room, we would die.)




What about Objectivity/Truth?


On Brandom’s pragmatic approach, how can normative assessments be rich enough to transcend the attitudes of practitioners?  How can two sentences have the same assertibility conditions but different truth conditions?

      Key idea:  Use the distinction between commitments and entitlements to distinguish two kinds of normative status:  subjective appraisal (justified) from objective appraisal (true). 


      1. “The swatch is red.”

      2.  “The claim that the swatch is red is properly assertible by me now.”

      3.  “I do not exist.”

      4.  “Rational beings never evolved.”




Brandom’s Semantics and Epistemology are Parochial, Based on Community Norms



What is the result?

Justification and truth are understood as “social statuses, instituted by the attitudes of linguistic practitioners”(189).


Parochial norms of justification and truth:  Consider an analogy to the rules of baseball.  They could have been different.  Could the rules of propriety for “justified” and “true” have been different?


For Brandom, truth and justification conditions are established by linguistic norms—conventions.  What was BonJour’s objection to the explanation of analyticity as truth by convention?  Could the conventions have been different?



Talbott’s Alternative Universalist Epistemology

Implicit vs. Explicit Sensitivity


1.  Universal concepts of Knowledge, Justification, and Truth

      Explicit Sensitivity/Explicit A Priori Insight Model:  Logical analyses of these concepts are the product of a priori insight into necessary truth

      Implicit Sensitivity Model:  Logical analyses of these concepts are the product of the imagination’s ability to conceive of possibilities.  Tentatively proposed necessary truths are those to which no one has been able to imagine a counterexample.  Logical analyses require us to be able to imagine a wide variety of cases and to make reliable judgments about particular actual and hypothetical cases. 

      (a) Knowledge à True Belief?

      (b) Knowledge à Truth?


2.  Universal principles of reasoning (e.g., mistake-correcting reasoning).

      Explicit Sensitivity/Explicit A Priori Insight Model:  Reasoning involves direct a priori insight into the principles of rational belief change, or at least, direct a priori insight into logical relations between premises and conclusion.  
















      Mistake correcting reasoning is an illustration of equilibrium reasoning, because it is non-monotonic (and thus non-inferential).  It involves judgments about what makes the most sense to believe. 


Given the information:

      (1) My sister Madeline’s child is named Ryan.


      Your preliminary conclusion:  Ryan is male.


Given the additional information:

      (2) Ryan is my niece.


      Your conclusion:  Ryan is female.


This is an example of non-monotonic reasoning, because it involves giving up a belief.  What are the principles that govern this kind of reasoning?


      Implicit Sensitivity Model:  We do not have a priori insight into such principles, we must discover them.  How could we discover them?  If there are universal principles of reasoning that our good reasoning is implicitly sensitive to, then we might discover those principles by attempting to find principles that would explain the difference between our good reasoning and our bad reasoning.  This requires us to be able to imagine a wide variety of cases of good and bad reasoning, to make reliable judgments about particular actual and hypothetical cases of good and bad reasoning, and to be able to formulate principles that would explain the difference.  If our judgments about good and bad reasoning are reliable enough, the principles that explain the difference between our good and bad reasoning will approximate the true principles of good reasoning. 


(Implicit Sensitivity) Agent A's coming to believe that p as a result of a transition from B1 and E to B2 is rational « There is a principle of rational belief change P which is such that agent A's coming to believe that p is appropriately implicitly sensitive to the application of principle P to the rational beliefs in B1 and A's experience E.















      (Test for Appropriate Implicit Sensitivity) Test for whether an agent A's acquisition of belief that p as a result of a transition from B1 to B2 given experience E is appropriately sensitive to the application of principle P to A's rational beliefs in B and to A’s experience E:  (a) If Principle P had not licensed A's believing that p on the basis of the rational beliefs in B1 and experience E, A would not have come to believe p as a result of B1 and E;  (b) If Principle P had licensed A's believing that p on the basis of the rational beliefs in B1 and experience E, A would have come to believe p as a result of B1 and E.



Naturalism and Normative Anti-Realism


Normative = prescriptive/proscriptive—for example, what I should believe or I ought to believe or what it is rational to believe or what I am justified in believing.


Evaluative = expressing a positive or negative attitude—for example, a good way of forming beliefs.


Purely descriptive = Neither normative or evaluative.











Sharon Street’s Attempt to

Eliminate Norms from Epistemology


Consider an epistemic norm—e.g., Occam’s Razor (a norm of theoretical simplicity):

“Whenever we make a normative judgment, we regard that judgment as true . . . .  On the other hand, we view our normative judgments as subject to causal explanation, just like any other psychological state”(233).


Street’s “Happy Coincidence” Objection to Normative Realism:  It would seem to simply be a coincidence that the causes of our normative judgments would have made them coincide with the true epistemic norms. 



To avoid having to believe in the “happy coincidence, the normative realist is committed to: 


Tracking account:  Our beliefs about epistemic norms track objective epistemic norms—that is, if our normative belief were not true, we (probably) would not believe it.


But the tracking account is scientifically unacceptable (234), because there is a better (theoretically simpler) explanation of our normative beliefs:


      Adaptive Link account:  Having true beliefs was advantageous for survival and reproduction (235).


      “The adaptive link account, I claim, defeats the tracking account hands down as a scientific explanation . . . .”(236).






The Problem of Self-Referential Inconsistency for Street’s Argument


Her argument depends on epistemic norms (e.g., a norm of theoretical simplicity).  If there are no true epistemic norms, how can she make her argument?


Street’s solution to the problem:  Replace all epistemic norms with purely descriptive statements involving probabilities.













Street’s Translation of the Norm of

Enumerative Induction

into Purely Descriptive, Probabilistic Terms


(EI-1) Norm of Enumerative Induction:  “The fact that n/m of the observed X’s have been Y is a reason to believe that n/m of the unobserved X’s are Y.”


(PD-1) Purely Descriptive Replacement for the Norm of Enumerative Induction:  The fact that n/m of the observed X’s have been Y makes it more likely that n/m of the unobserved X’s are Y.










But evolution selects for adaptive behavior, not for true explanations.  There are many purely descriptive replacements for the norm of enumerative induction that would have produced exactly the same behavior—for example:


(PD-2) The fact that n/m of the observed X’s have been Y is a reason to believe that n/m of the X’s observed in the future (but not those that never will be observed) are Y.


(PD-3) The fact that n/m of the observed X’s have been Y makes it more likely that God has arranged things so that n/m of the X’s observed in the future will be Y.








The ‘Happy Coincidence’ Objection

to Street


      There is no evolutionary explanation for why any theory about events millions (or billions) of years in the past would be likely to be true (including evolutionary theory).

      There is no evolutionary explanation for why any theory about objects at the molecular level (such as DNA) would be likely to be true (evolutionary theory depends on genetic transmission at the molecular level).

      Therefore, it could only be a happy coincidence if Street’s belief in evolutionary theory were true.

      But if it is irrational to believe in happy coincidences, then it is irrational for Street to believe in evolutionary theory.  Without evolutionary theory, her argument for normative anti-realism collapses.





      How could evolution have made human beings sensitive to universal principles of rational belief change?  By making us good learners.


      There are universal principles for being a good learner.  Evolution made us good enough learners about ordinary sized physical objects and the very same principles apply to theoretical beliefs about the distant past or the very small.  Our belief transitions are sensitive to those principles.














A more technical version of this answer: 


      The early Nozick’s surprising reply (before his “Copernican Revolution”):

Let R be the rational relation between the set of propositions B1 and the set of propositions B2.  Evolution could have selected for belief transitions that mirror rational relations.  Let b1 be the initial neurophysiolgical state of subject S and b2 be the final neurophysiological state of subject S.  Perhaps part of the explanation of subject S’s being caused to believe b2 on the basis of b1 is that b1 corresponds to B1 and B2 corresponds to b2 and B1 stands in R to B2. 




Nozick’s Test:  If B2 did not stand in relation R to B1, then it would not have been the case that both S is in b1 and b1 causes b2. 

Talbott’s Generalization of Nozick’s Idea:  Consider certain circumstances in which neurophysiological state b2 is the causal result of neurophysiological state b1 and neurological input e.  Where b1 corresponds to set of beliefs B1, e corresponds to a specification of experiential input E, b2 corresponds to set of beliefs B2, and the Principle of Rational Belief Change P licenses the transition from B1 and E to B2: 













(Generalized TIS) If P had not licensed the transition from B1 and E to B2, then it would not have been the case both that S was in state b1 and that b1 and e caused b2 in S (in these circumstances); and if P had licensed the transition from B1 and E to B2, then it would have been the case that b1 and e caused b2 in S (in these circumstances). 


The surprising possibility:  In selecting for adaptive behavior, evolution made us sensitive to rational relations among beliefs and experiential input.