Analyzing Knowledge
Analyzing Knowledge
We are familiar with Gettier problems, which bring out the insufficiency of the classic analysis of knowledge as true justified belief. But there are other problems with that analysis, centering on circularity. They question the pretensions of such an analysis to provide conditions that don’t presuppose the concept of knowledge but together add up to it. The classic analysis is at best misleading about the constitutive structure of the concept of knowledge, its inner constituents. The first problem concerns the concept of belief, usually introduced without much fanfare and little reflection. Thus, we are baldly told that if x knows that p, then x believes that p—while the latter does not entail the former. But what is it to believe that p? Isn’t it to take oneself to know that p? If I believe that the Battle of Hastings took place in 1066, I take myself to know that—and in this case I am right so to take myself. In other words, I believe I know it—or else I wouldn’t believe it. Thus, the concept of knowledge enters the concept of belief. If we were interested in defining belief, we might well offer: x believes that p if and only if x believes that he knows that p. True, we are using belief as part of its own definition, but the definition is informative and not viciously circular. We might also try to eliminate the concept of belief from it and replace it with something less repetitious (taking oneself to believe, or a disposition to assent, or brain-based functional role). In any case, the definition seems perfectly correct, and it invokes the concept of knowledge—the very concept we are supposed to be analyzing. What else is belief but an attempt at knowledge? How could you have the concept of belief and not realize that belief is linked to knowledge? Belief is would-be knowledge. Belief doesn’t entail actual knowledge, to be sure, but it alludes to that concept—it entails knowledge aspirations. The classic analysis only looks non-circular because we don’t ask ourselves how belief is to be understood, i.e., what it is. It is hardly a satisfactory entry point in the quest to specify what adds up to knowledge but doesn’t presuppose it.
The same question recurs with respect to the next two conditions—truth and justification. What is justification? Isn’t it precisely information that leads to knowledge, or is apt to lead to knowledge in favorable conditions? You can’t have the concept of justification and not see that justification leads to knowledge; no one could possess that concept and have never heard of knowledge. At any rate, we need an argument to show that such a thing is possible, or else circularity looms. Thirdly, the concept of truth is also inextricably linked to the concept of knowledge: truth is what the search for knowledge is the search for. How could you have a grasp the concept of truth and not know that truth is linked to knowledge in this way? The circle of concepts is too tight for that. You couldn’t learn what knowledge is by consulting the classic definition, because you would already need to have the concept in order to understand the definition. Truth and justification don’t individually entail knowledge, to be sure, but they incorporate the concept at one remove. The classic theory is not wrong exactly, but it fails to provide the kind of illumination advertised on its behalf. The concept of knowledge doesn’t pop out of the three conditions like the proverbial rabbit from a hat, because it already lurks close to the surface of the concepts used to define it, particularly belief. In the matter of conceptual priority, the concept of knowledge seems to come first—as primitive not constructed (as has often been contended).
These reflections encourage a reformulation of the classic analysis. This reformulation doesn’t dispose of the circularity problems, but it does simplify the intent of the classical analysis. It brings out what is really going on when someone knows something propositionally. Let’s say that x knows that p if and only if (1) x believes that he has a true justified belief that p (the belief condition), and (2) this belief is true (the truth condition). For he knows what knowledge is (true justified belief) and it is true that he is in the state in question. Intuitively, he believes he is in a state of knowledge and he actually is. The classic analysis analyzes his belief according to its own theory, incorporating this into the belief condition, and then it says that that belief is true. If it is correct in its analysis, then x will (perhaps tacitly) know what knowledge is, and then it only remains to specify that the conditions in question are actually satisfied. Knowledge emerges as a combination of belief and truth, just as we might have expected; but the belief turns out to have more structure than we might have supposed. We upload the analysis into the belief component, so to speak. This is really what is going on when a person knows something: he has a belief about his epistemic status and that belief is true. He doesn’t just have a belief in the proposition in question (e.g., that the Battle of Hastings took place in 1066); he also has a belief about this belief, to the effect that it is true and justified. This gives us a more accurate and illuminating picture of what knowledge entails—at any rate, of the kind of knowledge in question (rational, reflective). The person doesn’t just believe the first-order proposition, but also has other beliefs of an epistemological nature, concerning his entitlement to believe. The structure of the situation is made more apparent in this reformulation, though the substance is much the same (knowledge as true justified belief). The knower doesn’t merely have the belief in the first-order proposition; he also has various attitudes about his state of belief. Animals, presumably, don’t have such attitudes, so according to this analysis they don’t have knowledge (in the way we do at least). To have knowledge in the way a typical adult human does, you need to have the attitudes in question—which is the normal condition of human knowers. Knowledge is thus more richly structured cognitively under this formulation than under the classic formulation.[1]
[1] Here is a somewhat paradoxical result if Gettier is right and no one has an adequate analysis of knowledge: it will not be sufficient for knowledge that the subject’s belief about his epistemic state is true, since the true justified belief analysis is not sufficient. Accordingly, no one ever knows anything, or anything for which a Gettier case can be produced. In order to know, you need the correct analysis of knowledge; but according to Gettier no one knows the correct analysis, so no one knows anything. The strong belief condition rules out the possibility of knowledge, apparently. You can’t know without knowing what knowing is, but we don’t know that. Yet the reformulated classic analysis seems very plausible. One way out would be to say that we do know the correct Gettier-proof analysis, namely that knowledge is non-accidentally true justified belief—or whatever your favorite answer to Gettier is. This is an interesting wrinkle on the problem.
As someone deeply influenced by your earlier work, particularly your contributions to reliability theories, I find myself intrigued here by what seems to be a revisiting of the classical tripartite definition of knowledge. Your earlier accounts of knowledge as the upshot of a reliable discriminative capacity – emphasising the ability to ‘tell whether’ – have always struck me as profoundly illuminating, particularly in their ability to sidestep some of the pitfalls of the JTB framework, such as the circularity mentioned here.
So I’m curious about your motivation for revisiting the classical analysis in this recent essay. While the essay does an excellent job of exposing the conceptual entanglements within the JTB account, it appears, at least superficially, to contrast with the direction of your earlier work on reliability-based approaches, which offered a broader, more functional understanding of knowledge.
I wonder if your recent essay represents a deliberate return to foundational debates for purposes of critique or clarification, or perhaps it serves to recontextualise reliability theories within the broader epistemological landscape? Or might you be exploring a new philosophical direction that potentially brings these two frameworks into closer alignment?
Of course, I recognise that you understand these issues far better than I do and that what appears to me as a regression is undoubtedly more nuanced and intentional than I presently appreciate. In your earlier work, you acknowledge how, even with ‘reliability’ accounts, circularity threatens but you tend to favour a more “relaxed” approach, in terms of conceptual liaisons.
Thank you again for your continued contributions to philosophy, which have been a source of great inspiration and intellectual growth for me. I look forward to reading more of your work in the future.
It’s really just a matter of new ideas occurring to me. But I’m glad you have found my work illuminating. There is a systematic effort to downgrade it these days.
It’s disheartening, to say the least, to see how trivial irrelevances can sometimes overshadow the substantive merit of one’s ideas. That said, your work remains a crucial reference point for those of us who appreciate the value of clarity, rigour, and originality in philosophy.
I can’t disagree with you on that. How many current American philosophers would be willing to say as much?
On musical knowledge: reducing Ravel’s orchestral masterpiece ‘La Valse’ for solo piano, particularly as an alternative to his own reduction, was for me an exercise in a unique form of knowledge that transcends the domain of propositional truth. This process required a synthesis of theoretical understanding, practical skill, and aesthetic judgment. Unlike propositional knowledge, which concerns the distinguishing of truth and falsity, the knowledge I employed operated in the realms of interpretation and creativity, where correctness is not absolute or ‘binary’ but contingent upon my evaluations and priorities as an arranger.
In deciding to depart from Ravel’s reduction, I exercised what I can only describe as interpretative confidence – a belief that my understanding of the piece’s essence and my ability to adapt it for the piano could, in this context, surpass the composer’s own solution. While this may seem audacious, I believe it underscores an important philosophical point: in (musical) art, knowledge is not necessarily tied to authority or truth. Instead, it is warranted by an interpreter’s capacity to discern and execute choices that align with a convincing realisation of the work’s structure, texture, and emotional depth. For me, this autonomy is central to the evolving nature of musical performance and composition.
The knowledge I applied during this process manifested itself in my ability to make discriminative judgments – deciding what elements of the orchestration to prioritise, what textures to preserve, and how to convey the character of the piece within the constraints of the piano. These judgments were not about propositional truth but about achieving artistic coherence, effectiveness, and expressivity. Creating my alternative reduction was thus an application of know-how; practical knowledge that cannot be reduced to definitive rules or propositional statements.
By challenging Ravel’s original reduction, I highlighted for myself the value inherent in artistic knowledge. The inadequacy I perceived in his version made me realise that knowledge in the arts is not static or definitive but open to reinterpretation. Far from undermining his authority, my reinterpretation felt like an act of dialogue with his work, reshaping its legacy and asserting my role as a creative agent.
Ultimately, my reduction of ‘La Valse’ underscored the richness of musical knowledge as a dynamic interplay of theory, practice, and judgment. For me, this process demonstrated how art can transcend the constraints of truth-based epistemology, offering a space where knowledge is interpretative and personal. It was not merely an exercise in technical skill but a philosophical act of reimagining the boundaries of authority, tradition, and creativity.
That all sounds very plausible to me. I would say Lennon had greater musical knowledge, in this capacious sense, than McCartney.
The comparison is fascinating. Lennon’s bold creativity aligns with the ‘capacious’ knowledge I described – dynamic and interpretative – while McCartney’s strength lies in more technical territory.
Both are invaluable but Lennon’s approach better reflects artistic knowledge as evolving and personal (idiosyncratic, even).
Thank you for this enriching perspective.
Lennon had incredible musical taste; McCartney’s taste could be terrible (have you ever heard his Christmas song?).
Christmas is perhaps the season for novelty songs, and McCartney certainly embraced that (three times, all gloriously naff). And perhaps Lennon’s ‘Merry Christmas (War Is Over)’ is an exception, affording a touch of genius even in festive indulgence. This all reminds me of my own less-than-genius effort, ‘Christmas in Bognor,’ penned some 30 years ago. Yours is potentially better than McCartney’s, depending perhaps on the music (which I was struggling to imagine) Novelty has its place, don’t you think?
I haven’t even thought about the music, which may change the lyrics and structure somewhat; it’s a sketch for a song. All songs should have novelty.
I wonder if there is any AI that can provide a melody if you feed in the lyrics.
Sure, there’s OpenAI’s ‘MuseNet’, ‘AIVA’ (Artificial Intelligence Virtual Artist), ‘Amper Music’, ‘Soundraw’ and ‘Boomy’, all of which can assist.
But, of course, they all merely recombine patterns, having no evaluative capacity.
So they cannot transcend their algorithmic constraints.
Human creativity is also combinatorial but it isn’t necessarily algorithmic; it resists strict codification.
But they could suggest possibilities to me that I could creatively tweak.
Aye, for sure.
That’s what they’re good for (provided you know what you’re doing).