From PhilPapers forum Epistemology:

2011-10-06
Different kinds of knowledge have different justifications and definitions
I share the position (with Tristan Cunha) that Gettier has sufficiently refuted the JTB model, but for very different reasons.  I will return to these reasons below, but should first state that in my view an error is made in the way the question is introduced.  The "Gettier problem", if he's correct, is not Gettier's, but is a problem with the JTB version of the general correspondence model of knowledge which he strives to refute.  It functions roughly in the same way as Kant's dynamical Antinomies do for the claims Reason makes on the Understanding where unconditioned by possible experience:  to reliably constrain comprehensive claims.  Accordingly, the associated problem is not to repair the problematic criterion (as introduced by Zhuang on 2010-01-20 and attempted also by Tioa on 2010-01-30), but to look for another, if any exist, such as, for example, a "genuiness" criterion, where truth is not determined by correspondence to a knower, but rather a special variety of "being" a knower.  

With respect to my view of the results of the argument, it shows that some propositional contents of given beliefs could be true (correspond correctly under strict bi-valence) without being justified by any relationship those contents have to the belief-grounds by which those contents are held or asserted.  I interpret this part of the argument to be brought out by Zhuang in his 2010-02-12 response, annoting the truncation pointed out of a claim's truth from its tertiary grounds, so that justification of a given belief can have at most a mere accidental relation to the objective truth of its contents.  Therefore my support for Gettier's refutation is based on the radical disjunction of sufficient justification from the bi-valence of truth-claims, restoring a more cautious version of Hume's criteria of consistency of relations between habits of expectations and "bundles of perceptions", to repeat the famous phrase.  This in turn is by default a recommendation for a non-correspondence model altogether, where the relation between belief-justification and truth-claim-correctness is not accidental, but identical in individual cases, such where one makes an assertion about a "true work of art", for example.  

Chunha's attempt to repair the model by 1) collapsing appeals to authority into observations in general, as one variety thereof, and 2) subsequently distinguishing between observational and technical claim-contents (i.e. objects defined by how they look and how they are used) does not directly address (for propostions) Gettier's basic point:  the bi-valence of claim-contents maintains no non-accidental relation to the belief-justification.  By Chunha's argument the unconditionality of the relation is restated at the cost of elimination of observational contents from possible knowledge.  And besides, the use/observation distinction does not hold up at more minute levels, when one is learning a new skill, for example.