I wonder if this suggestion is Wittgensteinian at heart or whether it just seeks an agreed upon justification methodology. That is, would it be incompatible for someone who held meaning is attached to private states to demand an agreed upon methodology as you have here. And contrawise, would it be non-Wittgensteinian to allow for subjectively based justifications? I would think not so long as the meaning was tied to use such that the community of speakers could follow how the term was used and engage in the practice.

So what this boils down to is how to avoid Gettier cases, which do seem to arise from reasonable evaluations based upon incomplete knowledge. Your idea seems straightforward: force a community based standard for what constitutes a justification to avoid poor reasoning and perhaps require deeper investigation before declaring "knowledge."

If you tell me you're coming to my house, I see a blue jeep coming toward my house, you own a blue jeep, I say I know you're on the way, and you then arrive moments later to my house, we can say that I had knowledge of your arrival of the JTB variety. But then we learn it wasn't your jeep I saw and you took the bus, now we have a broken J, and a Gettier problem.

If you mean to add to the J methodology a stricter confirmation of all facts to avoid sloppier individualized justifications, that could be a solution, but I ask why that invokes Wittgensteinian other than perhaps reference to community involvement, but, as noted, the community could still use the word justification to mean whatever it decided without concern for avoiding Gettier.

That is, Wittgenstein wouldn't care whether a term were more useful. He'd only insist it's meaning were derived from use. — Hanover

I do think the framing is Wittgensteinian, but not because it appeals to “community agreement” as if justification were whatever a group votes into existence. The Wittgensteinian point I'm borrowing is about grammar: what makes a claim of justification intelligible is that it's answerable to standards of correct and incorrect application, and those standards are exhibited in a practice, in how we check, correct, and withdraw claims when error signals appear.

On “private meaning,” I'd put it this way. A person can have private experiences, and can have subjective certainty, but if the meaning of the terms involved were tied only to private states, then the distinction between correct and incorrect use would collapse. You could still demand an “agreed methodology,” but it would be unstable, because there would be no shared criteria to tell whether the methodology was actually being followed or merely seemed to be. That's why, in my framework, justification is not a private experience. It is practice-governed standing, and it is “objective” in the modest sense that the criteria for support, error, defeat, and correction can, in principle, be stated and applied within the practice. This is not consensus, not social permission, and not institutional authority, it is answerability to criteria.

That doesn't make subjectively based justification illegitimate. It means that subjective support has to be connected to use and to criteria that others can follow. If I say “I see blue,” or “I remember,” those are first-person claims, but they still live inside practices with mistake-conditions and correction, misperception, lighting, memory distortion, and so on. I'm not excluding subjective sources. I'm saying that their justificatory standing depends on how they are embedded in standards of assessment.

Now to Gettier. I'm not trying to avoid Gettier by requiring stricter confirmation of all facts. That would be impossible and it would smuggle in an infallibilist demand. The point is different: Gettier cases arise because we treat “seems justified” as if it were the same as having justificatory standing. In your blue-jeep example, what fails is not simply that you lacked a further fact, it is that the apparent support was not connected in the right way to the truth-maker, and the route is lucky. The practice would normally treat that as a fragile inference, and it would tighten the standards when the stakes are higher.

My proposal is not “let the community define justification however it likes.” It's: if we are using the word “justification” at all, we are already committed to certain constraints, no false grounds, practice safety, and defeater sensitivity, because those constraints are built into how justificatory talk functions in our life. Wittgenstein would not tell us to adopt a more useful vocabulary, but he would help us see what our vocabulary already commits us to when we use it.

If you want a one sentence summary: the Wittgensteinian element is not communal voting, it is the insistence that justification has a grammar of correct use and correction, and once we make that explicit, many Gettier intuitions are revealed as cases where the support was only apparent.

I'm no engineer, but it might look something like the following:

No False Grounds (NFG) = “Are we building on bad inputs?”
This is your QA/QC point. It asks whether the data or the key assumptions are incorrect in a way that would make the conclusion questionable.

Examples: wrong sample, mishandled sample, wrong method, transcription error, the lab did not follow procedures, etc.

Practice Safety = “Is the method we used a safe, normal way to reach this kind of conclusion?”
This is closer to standard of practice. It is not perfection, it is “we used a route that usually catches mistakes.”

Examples: proper calibration, chain of custody, replication, using accepted modeling procedures, etc.

Defeater Screening = “Even if the data are good, is there something that would overturn the conclusion?”

This is the part that is easiest to miss, because it happens after you think you are done.
It is the deliberate search for “what would make this conclusion fail.”

Examples in your setting:

A different source could explain the same contaminant pattern.

A missing geological feature changes the direction of some flow.

Seasonal changes that would modify an important consideration.

Another dataset (borings, field observations, historical site use) conflicts with the story you are telling.

So in one line:

NFG: inputs are not false.

Practice Safety: the route to the conclusion is not fragile.

Defeater Screening: no overlooked “gotcha” would overturn the conclusion.

That is how your quality program maps into my epistemology. That's the best I can do not being an engineer. It's just a matter of getting use to the procedure. Engineering has these procedures built into their conclusions.

I'll probably start a separate thread on that subject Tom. I'm not going to get into this subject here, but later in another thread. I'll just say this, most of the testimonial evidence is secondhand (hearsay), so by definition it's weak.

Isn't the annunciation of knowledge itself bound to the character of a localised hermeneutic. Do you give the least weight to individual or subjective testimony? Where is the rationale for weighted significance in your system for each or a combination of what you term, 'routes'? — Alexander Hine

Yes, the annunciation of knowledge is always situated in a local hermeneutic, a language, a practice, a way of drawing distinctions. I'm not trying to deny that. My point is that this doesn't reduce justification to “mere interpretation,” because within a practice there are criteria for correct and incorrect application, there are recognized mistake-conditions, and there are ways of correcting ourselves when the practice throws up error. The hermeneutic is real, but it isn't the whole story.

On individual or subjective testimony, I do give it weight. Testimony is one of the primary routes by which we acquire knowledge, and that includes first-person reports. The question isn't whether the report is subjective, it's how it stands within the standards that govern testimonial support: provenance, competence, independence, convergence, and defeater sensitivity. A single report is rarely self-authenticating, but it can still carry justificatory standing, especially when it's consistent, detailed, and later supported by independent lines of check.

As for weighting the routes, I'm not assigning a fixed hierarchy. I'm saying that the weight is determined by the case. In a given context we ask: which route is actually doing the work, what would count as a mistake in this domain, what would count as a defeater, and how strong are the correction mechanisms that are available. Then we look for convergence across routes, because that's often what turns a fragile support into stable standing. So the rationale for weight isn't that one route always dominates, but that different practices and different questions demand different standards, and the guardrails, No False Grounds, Practice Safety, and Defeater Screening, discipline whatever routes are in play.

You mean to elucidate for this audience that your project is a taxonomy of scientific method. — Alexander Hine

Not quite. What I am offering is a taxonomy of routes of justification that operate across many practices: testimony, logic, sensory experience, linguistic training, and pure logic in a boundary-setting role. Science is one prominent domain where these routes are integrated and disciplined by unusually strong correction mechanisms, but the taxonomy is not confined to science, and it is not meant to reduce every kind of knowing to scientific procedure.

The purpose is practical: when someone claims knowledge, I want to be able to ask, which route is doing the work here, what standards govern it in that domain, what would count as a mistake or defeater, and do the guardrails hold. That applies to science, but it also applies to ordinary life, history, law, engineering, and philosophy when philosophy is making knowledge claims rather than offering a mere stance.

If you want a quick check, a lot of what I call “knowledge” is acquired by testimony and linguistic training long before anyone does anything recognizably scientific.

I'm currently writing a book Why Christianity Fails using this epistemic model. Specifically, I analyze the testimonial evidence for the resurrection and demonstrate the weakness of the evidence.

From my paper:

Much of the contemporary discussion treats Gettier’s paper as showing that JTB is insufficient. I do not think this is the right lesson. The examples do not undermine the model itself. They depend on a confusion between what looks justified on the surface and what is genuinely justified within a practice. Once we attend to the structure of justification, including its graded and fallible character, it becomes clear that these cases fail to satisfy the justification condition in the first place. They rest on false grounds or on a lack of the relevant conceptual competence, and so they fall outside the classical model rather than threatening it. Seen in this way, Gettier does not overturn JTB; it signals the need to make explicit features of justification that the classical formulation left implicit. That is the task taken up by JTB+U in the sections that follow.

Worked Gettier example (diagnostic use). Consider the familiar “ten coins” case. Smith has strong evidence that Jones will get the job, and Smith has counted ten coins in Jones’s pocket. Smith forms the belief, “The man who will get the job has ten coins in his pocket,” by straightforward logical inference from what he takes himself to know. Unknown to Smith, Jones will not get the job. Smith will get the job, and Smith also happens to have ten coins in his own pocket. The belief is true, and it can look well supported, but it does not have the standing required for knowledge.

What fails is not truth, and not belief, but justification. The support Smith relies on depends on what is not the case, namely that Jones will get the job, and this triggers No False Grounds. One can say that Smith’s inference is valid, but validity is not enough, because justification is not merely a logical relation among propositions. It is a standing within a practice, fixed by public criteria that settle what counts as competent support in the context. The same case also brings Practice Safety into view. Smith stumbles into the truth by luck. In ordinary situations where the evidence is similar, he would draw the same conclusion, yet it would be false, so the belief is not practice safe. Defeater screening makes the point plain: once it is determined that Jones may not get the job, the belief loses its standing, and the only repair is to replace the faulty ground. Gettier does not refute JTB, it corrects a picture of justification as a private sense of assurance or a merely formal inference, rather than a public standing fixed by our epistemic practice.

Many acknowledge this, but then when pushed will only rely on science as if it's really the only method/s that counts. This is a confusion even among scientists. The problem is that most people (including scientists) don't have a good epistemology.

Science is not a single justificatory route that replaces the others. It is a practice that braids them together and then tightens the standards of correction.

Testimony: journals, lab notes, instrument reports, datasets, expert consensus, methodological inheritance.

Logic: inference, statistical reasoning, model selection, prediction, and constraint.

Sensory experience: observation, measurement, and interaction with the world through instruments.

Linguistic training: learning how to use the concepts correctly, what counts as a valid operational definition, what counts as a proper classification, what counts as a mistake in the domain.

Pure logic (boundary-setting): coherence constraints, definitional entailments, and the exposure of category mistakes.

Science is distinctive because it tends to force convergence by building systematic error detection into the practice. But the justificatory work still flows through the same routes. That is why it is a mistake to treat “science” as the only path to knowledge, and also a mistake to treat testimony as automatically inferior. The real question is the quality of the route in the case at hand, and whether the guardrails hold.

So you are borrowing from the type of standards that a scientific peer reviewed rationalism would apply to a systemic process philosophy? — Alexander Hine

In my framework, any proposed “method of justification” will usually be describable as a combination of the five routes I listed: Testimony, Logic (inductive and deductive), Sensory experience, Linguistic training, and Pure logic (boundary-setting only). The list is non-exhaustive in the sense that it doesn't pretend to capture every nuance of method, but it is meant to be covering in the sense that methods are built out of these elements, often in combination.

So, when someone proposes a new method, my first move is not to reject it, but to ask: which routes are actually doing the work here, and which guardrails are supposed to discipline them. Many disagreements then become clearer, because they turn out to be disagreements about which route is primary in the case, what the relevant mistake-conditions are, or which defeaters are being ignored.

If you think you have a method that does not pass through any of these routes, I would be interested to see it, but I suspect that in most cases what looks like a sixth method is really a composite that hasn't yet been analyzed under one or more of the methods I've outlined.

When I use the word “justification,” I am not talking about something private, a feeling of confidence, or a mere report of how things seem from a subjective point of view. I mean justificatory standing, the sort of standing a belief has when it is supported by the standards that govern a practice, standards for what counts as evidence, what counts as error, and what counts as correction.
— Sam26

So you mean Doxa? — Alexander Hine

Not doxa in the pejorative sense of mere opinion. I mean the normative standing a belief has when it is entitled by the standards of evidence and correction that govern a practice.

Here is something I stole from a post I made a few years ago.

A site conceptual model is just a description, image of the site which lays out all the information gathered during the investigations. To me, the most useful way of presenting a SCM is visually, using figures. Data tables are also needed. There will also be calculations e.g. groundwater flow direction and velocity, contaminant degradation rates, averages. On the figures, you can show the locations of the sources of the contamination and how it has moved and is presently distributed across the site. You can also show the expected distribution of contamination in the future based on groundwater and fate and transport modelling. You can also show the locations of existing and potential human and environmental receptors.

Typical data points include boring logs; analytical results of soil, groundwater, and sediment samples; visual observation of site conditions; topographic and bathymetric surveys; geophysical surveys; and wetland surveys. Going deeper, there are assumptions associated with laboratory analytical methods. Which in particular are you talking about? — T Clark

This is what I had in mind. A site conceptual model is a perfect case of knowledge that is not best expressed as a single proposition, but as an integrated representation: figures, tables, calculations, assumptions, and forecasts. In my terms, the unit being assessed is not one sentence but a model with correctness conditions, it can be more or less accurate, it can succeed or fail under error signals, and it can be revised when it runs into defeaters.

So, when I talk about justification here, I am not asking you to apply an engineering standard to a bare declarative sentence. I am asking how the SCM earns and keeps justificatory standing in the practice. Your description already points to the routes: sensory observation at the site, measurement and sampling, inference and modelling, testimony in the form of reports and lab results, and linguistic training in the way standards and classifications are applied. The important question is how those routes are disciplined.

That is where the guardrails map cleanly:

No False Grounds: what would count as a false ground in the SCM, a faulty assumption or input that is doing decisive work, for example a mistaken stratigraphic interpretation, a mislocated source term, or an analytical artifact that propagates through the map of contamination.

Practice Safety: what makes the SCM robust rather than lucky, for example triangulation across independent data types, sensitivity checks, conservative assumptions where appropriate, and repeated checks that would expose a fragile inference.

Defeater Screening: what kinds of findings would force revision, for example a new boring log that contradicts the stratigraphy, a plume boundary that violates the predicted hydraulic gradient, or receptor evidence inconsistent with the proposed pathway.

On your last question, I am not asking about one laboratory method in the abstract. I mean the assumptions that bear the weight in the overall chain that supports the SCM.

If justification is a standing within a practice, then understanding is the competence by which a person can genuinely participate in that practice, not merely mimic its conclusions.
— Sam26

I think I've grasped how you use "understanding" here, and why it isn't a fourth criterion for knowledge, but rather an attempt to clarify what justification actually entails. At this point, an example would be helpful. The question of what it means to "mimic the conclusions" is central, I think. You write:

Someone can hold a true belief and even cite a correct supporting data, while still failing to grasp what that support is doing, what would count against it, what would defeat it, and what would count as a relevant correction. In that situation the belief can look justified from the outside, and even to the person themselves, but the justificatory standing is fragile, because the person does not reliably track the mistake-conditions that the practice treats as decisive.
— Sam26

As an example, I picture a student who writes a paper on a topic in science; the paper describes a true belief which the student holds, and cites all the correct data. Why is the student only mimicking the conclusions? Because their understanding of why the data provides a justification has to involve a simultaneous understanding of the conditions under which they wouldn't -- the defeaters, in other words. If the student lacks this understanding, their claim to justification is shaky, and probably false.

Does that sound right? OK, here's the question: If the belief "looks justified" both from the outside (publicly) and to the person themselves (privately), how should we describe the process that will show us it's not justified? It seems as if a verification of understanding requires a further, dialogic process with the one who claims justification (and knowledge). And that's fine, but perhaps you should emphasize the need for this further step. I agree that it still doesn't make for a fourth criterion, but it does seem significantly different from the process we would engage in to learn a person's justifications, which, as you point out, can be merely cited or mimicked. Another homely example might be defending a thesis.

Maybe all of this is to say we can't "vet understanding" in the same way we can vet a proof, or even a proposition. The proof doesn't reply to our questions, but we do require the person to, otherwise we're not in a position to say whether the U part of JTB+U is present. This doesn't contradict your theory in the slightest, just elaborates it a bit, and puts it in a context of Habermasian "communicative action." — J

This is a strong reading of what I meant, and your student example captures the central point. “Mimicking the conclusions” is precisely the case where a person can reproduce the correct outputs, cite the right data, and sound fluent, while lacking a grasp of the mistake conditions, the relevant defeaters, and the revision pathways that the practice treats as decisive. In that situation, the belief can look justified, even to competent observers, because the surface marks of justification are present, but the standing is fragile because it's not anchored in the competence that makes those marks responsibly usable.

Your question about how we show that the belief is not justified is also right, and it helps to make explicit something I left implicit. In many domains we do not vet understanding by inspecting a static artifact alone, as if it were a completed proof. We vet it by exposing the claimant to the practice’s tests, especially its countercases. That often does require a dialogic dimension: questions, challenges, requests for boundary conditions, requests for what would count as defeating information, and requests for how the claim would be revised if those defeaters obtained.

But I want to put this carefully, so it does not look like an added criterion. The “further step” you describe is not a separate requirement piled onto justification, it's one of the ordinary ways a practice determines whether a person has justificatory standing or has only borrowed it. It is the difference between an utterance that happens to be correct and a competence that can carry that correctness across the relevant cases. In that sense, the dialogic process is a method of assessment, not an additional condition of knowledge.

It is also worth noting that the need for dialogue varies by context. Sometimes understanding can be vetted through performance without explicit conversation, for example by reliable error detection, appropriate revision under new data, or correct handling of nearby cases. In other settings the quickest test is indeed oral examination, thesis defense, or cross examination. Either way, the underlying point is the same: understanding is shown in how the claimant navigates defeaters, boundary conditions, and correction, not merely in the ability to cite supporting considerations.

So, I agree with your closing line as an elaboration: we cannot vet understanding in the same way we vet a proof considered as a static object. We vet it by putting the claimant into the space of questions and challenges that the practice treats as intelligible. That is compatible with my view, and I think it helps readers see that “public” does not mean “a pile of citations,” it means susceptibility to the practice’s checks, including dialogic ones when the case calls for it.

If you want a single sentence version of the answer: when a belief looks justified on the surface, the practice distinguishes genuine standing from mimicry by testing whether the person can track defeaters and revise under correction, and that test is often, though not always, dialogic.

I've read all seven of your chapters. Just for workability, I'm going to respond to each chapter separately. This may mean that what I have to say will be a bit disjointed. We'll see.

My biggest overall issue--JTB generally applies to propositions but most of the knowledge we have and use is not really expressible in that form. As an engineer, I usually talked about "conceptual models," which means an overall picture of the situation--in my case it was real estate properties and the soil and groundwater characteristics distributed across the site and at different depths. Models like that will generally be judged and justified as accurate rather than true. As I indicated, as I see it, the way we use knowledge on a daily basis tends to be more like how I've described it rather than just the truth of propositions.

This is highlighted by your discussion of the idea of standards of practice which are used to justify truth. In general, I think that's right, but how standards are applied under JTB (or JTB-U) is different from how various practices apply their standards. How do I apply an engineering standard to a simple declarative statement?

So, I worry that I am going to send your discussion off on a tangent. Now that you've seen some of the substance of my thoughts, should I continue? — T Clark

I don't think this is a tangent, I think it's exactly the kind of stress test that helps clarify my ideas.

On the first point, I agree that a great deal of what we rely on is not best described as a single proposition. Much of it is a competence, a grasp of a situation, a model, or a way of seeing how things hang together. In engineering, the object of assessment is often a conceptual model, and the operative question is whether the model is accurate, robust, and fit for use across the relevant conditions, not whether a sentence is true in isolation.

My reply is that this does not put the JTB family out of business, it forces a clarification of what “truth” and “justification” are doing. A model can be assessed for correctness in the world, it can succeed or fail, it can be refined under error signals, and it can be defeated by counterevidence. In other words, it has correctness conditions even if it's not naturally expressed as a single declarative statement. The “propositional” layer can be treated as a partial extraction from the model, for example, predictions, constraints, and consequences that can be checked. That is often how the model earns and keeps its standing.

On your second point, I agree that we should not imagine applying an engineering standard to an isolated declarative sentence as if that were the primary unit of knowledge. The better way to put it is that standards of practice govern the evaluation of the claim in its proper form, which may be a model, a procedure, a measurement protocol, or a forecast. When I say “public” or “practice-governed,” I mean that there are criteria for correct application, error, and correction that can be articulated and contested within the practice. Engineering seems like a textbook example of this, because the standards include calibration, measurement error, boundary conditions, sensitivity to assumptions, and the discipline of revising the model when it fails.

So, I would welcome you continuing, but with one focusing suggestion so we do not drift. When you respond to a post, pick one concrete engineering example of a conceptual model and say how it is justified in your sense. Then we can map it onto my vocabulary without forcing it into a single sentence: which route of justification is doing the work, what would count as a defeater, where No False Grounds shows up, and what “practice safety” looks like in that domain. If we can do that once, the “propositions versus models” worry will either dissolve or become precise enough to evaluate.

Optional sidebar: Hinges and Gödel, a structural parallel

A brief note for readers interested in foundations. In my paper I suggest a structural parallel between hinge certainties and Gödel style limits. The parallel is not evidential, and it is not a proof of anything in epistemology from mathematics. It is a comparison of structure.

In Gödel’s setting, once a formal system is rich enough, there are truths expressible within the system that cannot be proven by the system’s own resources, and consistency cannot be established from within in the strongest way one might want. The upshot is not that mathematics collapses, but that the practice operates with boundary conditions that are not resolved by the same methods the system makes possible.

In the hinge setting, justificatory practices also have stopping points. Certain things stand fast, not as conclusions of inquiry, but as the background that makes inquiry, doubt, evidence, and correction possible. The upshot is not that justification collapses, but that justification always operates within a framework whose role is not that of an ordinary claim awaiting ordinary support.

So the comparison is this: both domains exhibit limits on what can be achieved from within, and both continue rationally once those limits are acknowledged. That is all I mean by the parallel.

If you want to press on this, I would welcome it, but it will help to keep the debate focused on whether the analogy is illuminating rather than on technical details of Gödel’s proofs.

Much of this is already written out, which is why I can respond quickly sometimes.

Clarifications and terms (so we do not talk past each other)

Before continuing, I want to clarify a few terms that can easily be misunderstood.

What “+U” is and is not. In JTB+U, “Understanding” is not a new ingredient bolted onto JTB. It is a way of making explicit what justification already presupposes in ordinary epistemic practice, the competence to grasp what counts as support, what counts as error, and what would count as correction in the domain.

What I mean by “public.” When I say justification is public, I do not mean popular agreement or institutional permission. I mean that justification has criteria that can, in principle, be articulated, assessed, challenged, and corrected within a shared practice.

Practice Safety is not a demand for infallibility. Practice Safety means that the route by which a belief is held is not fragile or lucky with respect to the mistake conditions the practice recognizes. It is about tracking error signals in the domain, not about achieving certainty in the absolute sense.

What I mean by a defeater. A defeater is not merely disagreement. It is a consideration that, if true, would remove or weaken the support, or would show that the apparent support is misleading. Defeater screening is the discipline of identifying and facing such considerations rather than ignoring them.

Hinges are not reasons. Hinges are not hypotheses supported by evidence. They are background certainties that stand fast and make evidence, doubt, and correction possible in the first place.

Four uses of “certainty.” I distinguish subjective certainty (conviction), hinge certainty (bedrock), epistemic certainty (defeater resistant stability in practice), and absolute certainty (logical or moral necessity). When we disagree, it often helps to say which sense of certainty is in play.

Post 7 of 7:
Gettier, and why “seems justified” can mislead

Gettier cases are designed to make the traditional JTB account look inadequate. The familiar pattern is that a person has a true belief, and can cite what appears to be a justification, yet we hesitate to call it knowledge. The standard moral is that JTB is missing some extra condition. My paper argues that this moral is too quick, because it treats “justification” as if it were exhausted by having a supporting consideration that can be stated.

On the view I am developing, the key distinction is between a belief that looks justified, and a belief that has genuine justificatory standing within a practice. In Gettier style cases, the subject often has support that is either dependent on a false ground, or is insulated from the relevant mistake conditions, or is undermined by an undefeated defeater. In other words, the cases exploit a gap between seeming to meet the justification requirement and actually meeting it once the ordinary constraints on justification are made clear.

This is where the guardrails matter. If a belief depends on a false ground, No False Grounds blocks it. If the route is fragile and the belief is true by luck, Practice Safety blocks it. If there is relevant defeating information that has not been faced, Defeater Screening blocks it. The upshot is not that Gettier reveals a defect in JTB, but that Gettier is trading on an impoverished picture of justification, and once justification is properly described, the cases no longer force an additional condition.

This also brings the “+U” into focus. Understanding is not a decorative addition. It marks the competence by which a person can genuinely track what their support does, what would count against it, and what would require revision. A person can cite a reason and still be out of contact with those mistake conditions. When that happens, the belief can be true and can look justified, yet it lacks the stability we normally require for knowledge.

Upshot: Gettier cases do not show that knowledge needs a mysterious extra ingredient. They show that we should not confuse the appearance of justification with genuine justificatory standing.

Questions for critique:

Do you think Gettier cases still refute JTB even if we build in the guardrails and the “+U” clarification.

Is my diagnosis too dependent on relabeling the justification condition rather than answering the core intuition.

If you have a favorite Gettier case, post it and say which guardrail you think it slips past, if any.

Post 6 of 7:
Hinges and why justification has stopping points

At some point justificatory questions come to an end, not because inquiry has failed, but because the very practice of giving and asking for justification presupposes a background that is not itself justified. If you demand a justification for everything, including the conditions that make justification possible, you do not reach a deeper standard. You undermine the justification.

This is where hinge certainties come in. A hinge is not a hypothesis, not a conclusion, and not a belief held because we have evidence for it. It is something that stands fast in a practice, a bedrock commitment expressed in how we proceed, what we take for granted, what counts as doubt, and what counts as a mistake. Hinges are not the kind of things we arrive at by argument, but they are also not arbitrary. They belong to the inherited background against which reasons, evidence, and defeaters can have their force.

That matters for two reasons. First, it explains why the demand for ultimate proof is misguided. Proof and justification always operate within a framework, and the framework is not itself established by the same tools it makes possible. Second, it explains why skepticism so often feels powerful. Skeptical questions typically target hinges and treat them as if they were ordinary empirical claims. Then, when those hinges cannot be proven in the skeptic’s demanded way, skepticism concludes that nothing can be known. The mistake is grammatical. The skeptic is asking for a kind of justification that cannot apply to the role hinges play.

None of this licenses dogmatism. It is true that some hinges can shift as practices are repaired, methods change, or persistent error signals force a reorientation. But it is equally true that some hinges do not shift, at least not within anything we would still recognize as the same form of life. Their role is constitutive, they are part of what makes inquiry, correction, and assessment possible at all. Where a hinge does shift (e.g. we are objects separate from other objects), the change is usually not a matter of ordinary argument but a deeper reorganization of the practice itself.

Upshot: hinges are not additional reasons. They are the background that makes reasons and defeaters possible, and recognizing this prevents both regress and skeptical distortion.

Questions for critique:

Do you think hinges are real features of our epistemic life, or are they a philosopher’s invention to stop regress.

Does treating skeptical challenges as hinge confusion actually answer skepticism, or does it merely set it aside.

Can you name a hinge you think my framework relies on, and say whether you think it should stand fast or be challenged.

:up:

Post 5 of 7:
Five routes of justification (non exhaustive)

In the paper I lay out five primary routes by which beliefs commonly acquire justificatory standing. The point is not to rank them or claim that every case fits neatly into a single category. The point is to map the main ways we actually come to know, so that we can ask where a claim is getting its support, and what standards and error conditions belong to that route.

Testimony. Most of what we know comes from others, ordinary reports, books, videos, expert claims, historical records, and lived witness. Testimony can confer justificatory standing, but it has its own standards: credibility, independence, competence, convergence, and the absence of relevant defeating information.

Logic (inductive and deductive reasoning). We justify beliefs by inference, sometimes strictly, sometimes probabilistically. Here the relevant standards include valid form where appropriate, good inductive support, sensitivity to base rates, and the ability to identify where an inference is overextended.

Sensory experience. Experience is a central route of support in ordinary life. It has its own error conditions: illusion, distortion, poor conditions, and conflict with other well established checks. Sensory experience does not need to be perfect to justify, but it must be situated within the ordinary corrective practices that make perception reliable in the domain.

Linguistic training. Some things are “known” because we are trained into a practice, trained to use terms correctly, to recognize criteria, to follow rules, and to distinguish correct application from misuse. This route is often invisible because it is basic to how we learn the grammar of our concepts, but it is indispensable for explaining how justification becomes possible at all.

Pure logic (boundary setting only). There are limits that are not empirical discoveries but logical constraints, what is possible, what is coherent, what follows from definitions, what collapses into contradiction. This route does not supply new facts about the world. It sets boundaries, clarifies entailments, and exposes category mistakes.

This five route map is not meant to replace the earlier guardrails. The guardrails discipline justification. The routes describe where justification is coming from. In any concrete case, the question becomes: which route is in play, what are its standards, and do the guardrails hold.

Upshot: the routes give us a practical way to locate a claim in the space of support, standards, and error conditions, without turning epistemology into a single method.

Questions for critique:

Are these five routes a helpful map, or do you think the categories blur in a way that makes the list misleading.

Do you think “linguistic training” deserves to be a distinct route, or is it better treated as part of the background of the other routes.

Is my use of “pure logic” as boundary setting clear, or does it need a different label.

Post 4 of 7:
Three guardrails that discipline justification

If justification is a standing within a practice, it still needs discipline. Not every chain of support confers standing, and not every true belief that happens to be well supported counts as knowledge. In the paper I use three guardrails to mark common ways justification fails, even when a belief looks respectable.

No False Grounds (NFG). A belief cannot have genuine justificatory standing if the support it depends on is false, or if it is being carried by a false presupposition that is doing the real work. The point is simple: if the ground is false, whatever looks like support is a counterfeit support.

Practice Safety. A belief is practice safe when it is formed and maintained in a way that reliably tracks the mistake conditions recognized by the practice. This is not infallibility. It is the idea that the route by which the belief is held is not fragile, lucky, or insulated from the ordinary error signals that would count against it in a particular domain.

Defeater Screening. Even when the grounds look good, justificatory standing is undermined when there are live defeaters that have not been faced. A defeater is not merely a contrary opinion. It is a consideration that, if true, would remove or weaken the support, or would show that the apparent support is misleading. Defeater screening is the discipline of identifying and addressing such considerations, rather than ignoring them.

These are not meant as extra conditions stapled onto JTB. They are ways of making explicit the constraints that ordinary epistemic practice already applies when it distinguishes genuine support from luck, from illusion, and from rationalization.

Upshot: the guardrails do not add a new theory of knowledge, they articulate the failure modes that explain why “seeming justified” can diverge from genuine justificatory standing.

Questions for critique:

Do you think these guardrails capture real failure modes, or do they smuggle in something stronger than ordinary justification.

Is Practice Safety a useful idea, or does it collapse into defeater screening or into reliability talk.

Can you think of a case where a belief meets these guardrails and still intuitively fails to count as knowledge.

Yes, it makes sense for you to participate, and I think your pragmatic approach can fit naturally with mine, as long as we keep the different uses of “certainty” from sliding into each other.

I divide certainty into different uses:

Subjective certainty, conviction, how settled a claim feels.

Hinge certainty, the bedrock that stands fast and makes inquiry and doubt possible at all.

Epistemic certainty, defeater resistant stability in practice, the kind of standing we treat as enough for responsible action within a domain.

Absolute certainty, logical or moral necessity.

With that in view, your “adequate certainty about outcome” sounds closest to what I call epistemic certainty, not absolute certainty, and not merely subjective certainty. In other words, it is not infallibility, and it is not just confidence. It is a claim having the right kind of stability under the relevant checks, given the stakes and the standards of the practice you are operating in.

Where our approaches might meet is that both of us think justiification is disciplined by standards. Where we might differ is that I frame those standards in terms of justificatory standing within a practice, including what would count as a mistake, what would count as a defeater, and what would count as a responsible correction. Practical stakes can raise the bar, but they do so by tightening what counts as adequate support, not by demanding absolute certainty.

So yes, participate. If you want a clean point of contact, tell me what your “adequacy” standards are in a concrete case, and we can ask how they relate to defeater resistance, false grounds, and practice safety, and which sense of certainty they are aiming at.

Post 3 of 7:
What “Understanding” is doing in JTB+U

In JTB+U, “Understanding” is not a new mental ingredient added on top of truth, belief, and justification. It names a feature already at work in justification, the grasp of the concepts and inferential roles that make justificatory standards applicable at all. If justification is a standing within a practice, then understanding is the competence by which a person can genuinely participate in that practice, not merely mimic its conclusions.

Someone can hold a true belief and even cite a correct supporting data, while still failing to grasp what that support is doing, what would count against it, what would defeat it, and what would count as a relevant correction. In that situation the belief can look justified from the outside, and even to the person themselves, but the justificatory standing is fragile, because the person does not reliably track the mistake-conditions that the practice treats as decisive. Understanding, in this sense, is the operative structure of justification, the competence that makes justificatory standards genuinely operative rather than merely repeatable.

This is also why I resist treating “justification” as if it were only a list of supporting propositions. A list can be repeated by rote. Understanding is what makes the support more than a recitation, it is the ability to locate the claim within the space of reasons, objections, defeaters, and revisions that the practice recognizes. That is not infallibility, and it is not an impossible demand. It is simply the difference between having a standing and merely borrowing one.

Upshot: “+U” marks the competence that makes justificatory standards operative, it is not a separate add on.

Questions for critique:

Is this notion of understanding genuinely distinct from justification, or does it collapse into it.

Does tying understanding to error signals, defeaters, and correction make the account clearer, or does it over intellectualize ordinary knowing.

Can you think of a counterexample, a case where someone lacks this competence but still seems to have genuine justificatory standing.

Post 2 of 7:
What I mean by “justification”

When I use the word “justification,” I am not talking about something private, a feeling of confidence, or a mere report of how things seem from a subjective point of view. I mean justificatory standing, the sort of standing a belief has when it is supported by the standards that govern a practice, standards for what counts as evidence, what counts as error, and what counts as correction.

That is why I sometimes describe justification in terms of grammar (Wittgensteinian grammar). The point is not that knowledge is “only language,” but that the difference between being entitled and merely thinking one is entitled is built into how our practices work. We learn what it is to justify by learning how claims are checked, challenged, repaired, and sometimes withdrawn. Those norms are not optional decorations added after the fact. They are part of what makes the idea of justification intelligible.

This is also why I emphasize the public character of justification. “Public” here does not mean popular agreement or institutional permission. It means that justification has criteria that can, in principle, be articulated, assessed, and disputed within a shared practice. A belief can be held with sincerity and conviction and still fail to have justificatory standing.

Upshot: justification is a normative standing within a practice, not an inner endorsement, and that is the background against which the “+U” move makes sense.

Questions for critique:

Do you think justificatory standing can be explained without appeal to shared criteria, or does that collapse into a purely psychological picture.

Is my use of “grammar” illuminating here, or does it obscure what is really going on.

Does “public in this sense” capture what we need for justification, or does it leave out something essential.

You can comment as I go along, but many of the questions about what I mean by this or that will be explained in later posts.

The spine of my paper will be covered in seven posts.

1. What JTB+U is claiming

2. What I mean by justification as practice grammar

3. What “Understanding” is doing

4. The guardrails (No False Grounds, Practice Safety, Defeater Screening)

5. The five routes of justification

6. Hinges as non-epistemic background

7. Gettier diagnosis and upshot

Post #1
What JTB+U is claiming

My paper defends a practice-oriented refinement of the traditional JTB account. I call it JTB+U: “justified true belief” plus Understanding. My point is not to add a new requirement onto JTB as if we had discovered a fourth ingredient. My point is to make explicit something justification already presupposes in everyday epistemology, i.e., the competence to grasp what counts as support, what counts as error, and what would count as correction within a practice.

On this view, many familiar puzzles arise when we picture justification as something essentially private, an inward sense of being entitled (e.g. the use of know as an expression of a conviction), as if the fact that a belief feels well supported could stand in for the standards by which it is actually assessed. But in ordinary epistemic life, justificatory standing is not conferred by confidence or by an internal impression of rightness. It depends on the grammar of our practices, the criteria by which we count something as evidence, the ways we identify error, and the norms by which we correct it. When we bring that grammar into view, Gettier cases lose much of their force because many depend on a mismatch between seeming justified and having justificatory standing.

Upshot: I am not replacing JTB, I am strengthening it by making explicit the Understanding that is already doing quiet work inside justification.

Questions for critique:

Does the “+U” clarify anything real, or is it a relabeling.

Where would you draw the boundary between justification and understanding, if you think there is one.

What is your strongest reason to think Gettier still bites even after this move.

Morality from a secular position is necessarily subjective. — Ram

This premise isn't true. The key is that you said necessarily. To counter a necessity claim, I do not need to show that all morality is objective, or even that all harm is objective, I only need one counterexample. For example, suppose I cut off someone’s arm for no good reason. The harm in that case is not a matter of opinion or private feeling. It is publicly observable, i.e., objective: an arm on the ground, blood loss, shock, the screams of the one harmed, the reactions of witnesses, the lasting impairment. Anyone can see what has happened, and anyone can see that nothing about this depends on my personal preferences.

Now you might say, “Fine, the harm is objective, but calling it wrong is still subjective.” But that is exactly where the word necessarily overreaches. In ordinary moral judgment, severe harm functions as a public defeater: if you cannot give reasons that others can evaluate as sufficient, the act is not merely “disliked,” it is impermissible. You can reject that grammar if you want, but then you are no longer describing morality so much as evacuating it. So, at a minimum, this one case shows that a secular moral judgment can be anchored in objective features of the world and in publicly assessable justification, which is enough to refute the claim that secular morality is necessarily subjective.

I say this as someone who is not a secularist, but as someone who allows for an expanded metaphysics.

Hello everyone,

I'm writing to share that this will be my final post for a few months. I need a break.

Sam

My general impression of your narrative says, "You want to pair the metaphysics of knowledge relationships (p →q), as dynamically governed by an emergent and energetic inter-relation, viz., truth, with empirical experience. Dynamical, energetic identity transformations across space and time forming symmetries that conserve identity and support an enduring POV embody the living experience of truth. — ucarr

In other words, truth isn’t some hidden essence, it’s what happens when our justified beliefs line up with the facts of the world, or the way the world is. We test truth through shared practices (Wittgenstein's language games, which are governed by implicit rules), our forms of life, where we check, correct, and agree on what counts as evidence. In some cases, like science or mapping, truth can be pictured or measured, but even there it works only because we understand what the picture means and how it connects to reality. Understanding (JTB+U) isn’t optional; it’s what lets us tell genuine truth from lucky coincidence.

Formal theories of truth, like those used in logic, capture a structure but not its lived reality. They can show when a statement fits certain conditions but can’t explain how truth operates in lived reality, how it shapes belief, correction, and meaning. Truth, as we actually experience it, isn’t a Tarski formula (“p” is true iff p.) but a practice. There's a philosophical bridge between ontology and epistemology: the world has its own structure (what obtains), and we have our structures of reason, language, and justification. Truth is the point where those two orders (the world and epistemology) align.

In the framework I use, truth is not a metaphysical essence but a relation intrinsic to our practices of justification. To say a proposition is true is to claim that it holds up under the public criteria of a form of life, viz., that it connects belief with what obtains in the world. Truth marks the point where our language intersects with reality and is further illuminated by understanding: not merely that the world is as the proposition says, but that we can see how and why this is the case. The correspondence is real and, in some language-games, legitimately pictorial, e.g., where mapping, modeling, or measurement aim to reproduce structure or proportion. Yet even there, “picturing” works only because it is guided by understanding: without grasping how the representation functions, no degree of accuracy would amount to knowledge. Understanding is easily overlooked because it seems built in, but it is what allows us to apply the criteria of truth, to distinguish success from coincidence, evidence from echo. What makes a proposition true is the state of affairs that obtains; what makes that truth knowable is the grammar of our interaction with it, governed throughout by understanding. In this sense, truth is both discovered and articulated, anchored in reality and shown through our capacity to comprehend its order.

Formal definitions of truth, though indispensable in logic, leave this fuller picture out. Tarskian or semantic schemas (“‘p’ is true if and only if p”) capture the structure of truth but not its life. They specify conditions of equivalence but remain silent about how truth functions within inquiry, how it guides belief, sustains correction, and grounds public justification. Formal accounts strip truth of human context: they can model consistency but not meaning, accuracy but not understanding. What they describe is the form of truth’s operation, not its practice. Truth, as lived and recognized, is not a symbol in a metalanguage but what’s embodied in our forms of life (our language games), the point where the world’s order and our conceptual order momentarily coincide.

Can you name a few of those "forward-looking thinkers"? — Janus

There are those who view AI as an epistemic tool, something that extends, rather than replaces human inquiry. There's a long list of people who fit the bill. For example, Nick Bostrom and Luciano Floridi have been working on the conceptual implications of AI for ethics, cognition, and the philosophy of information. Vincent Müller and Mariarosaria Taddeo have been exploring how AI reshapes the logic of justification and responsibility in scientific reasoning. On the cognitive side, Joscha Bach treats AI systems as experimental models of mind, ways to probe the nature of understanding. Even researchers outside philosophy, in fields like computational linguistics and mathematical discovery, are beginning to treat AI as a genuine collaborator capable of generating new proofs and hypothesis.

Maybe we use books, dictionaries, philosophical papers, editors, and scientific discoveries to make us look smarter than we are. You see this all the time in forums, even without AI, so it's nothing new. Besides do you really care about the psychology of someone who's writing about what they think?

I won't comment on the political part of your post because I think we're very far apart. However, in the future I can see where humans will merge with AI, so we'll probably become one with machines, probably biological machines.

The objective in thinking for yourself is to take every idea you hear from others with a grain of salt, and to even question your own ideas constantly. — Harry Hindu

If you take every idea with a grain of salt, you’ll never move beyond hesitation. Critical thinking isn’t about doubting everything, it’s about knowing when doubt is justified. In logic, mathematics, or physics, for instance, constant suspicion would paralyze learning; you suspend doubt provisionally because the framework itself has earned trust through rigor.

In a philosophy forum, though, caution makes sense. Most participants lack grounding in epistemology, logic, or linguistic analysis, so what passes for argument is often just speculation dressed up as insight. Honestly, you could gain more from interacting with a well-trained AI than from sifting through most of what appears here, it would at least give you arguments that hold together.

Much of what all of us do is "parrot." Not many people can come up with an original idea to save their life.

It's already helped me expand my thinking on epistemology, and it gave me good ideas on my book. However, you do have to have prior knowledge because it does make mistakes. The next two iterations of ChatGPT and Grok 5 have a good chance to reach AGI.

AI models from OpenAI and Google DeepMind reached gold medal-level performance at the 2025 International Mathematical Olympiad (IMO), one of the most prestigious math competitions in the world. It's also better at diagnosing than many doctors. So, I don't know where you're getting your information.

I've come to see anything that is not based on rigorous analysis or scientific understanding as intellectual wankery—mental masturbation—and I have no problem with people enjoying that, but the idea that it is of any real significance is, for me, merely delusory. — Janus

Don't mistake the speculative misuse of ideas for the ideas themselves. AI is no longer in the realm of “mental masturbation,” it’s already reshaping science, mathematics, and even philosophy by generating proofs, modeling complex systems, and revealing previously inaccessible patterns of thought. To dismiss that as delusory is to confuse ignorance of a subject with the absence of rigor within it.

The irony is that the very kind of “rigorous analysis” you claim to prize is being accelerated by AI. The most forward-looking thinkers are not treating it as a toy but as a new instrument of inquiry, a tool that extends human reasoning rather than replacing it. Those who ignore this development are not guarding intellectual integrity; they’re opting out of the next phase of it.

Truth is an emergent feature of linguistic and conceptual frameworks; it depends on the existence of propositions and shared criteria of correctness.

The fact is that if you don't know what you're doing, the result will be a mess. I've used AI for programming before and you really have to guide it and pay close attention to everything it does and constantly question its decisions. — Jamal

I wholeheartedly agree with your assessment. I used AI extensively while writing my book on NDEs and my work on epistemology. It was helpful for editing and idea generation, but it also made frequent errors, enough that I often wondered if it was creating more work than it saved. You have to know the material well to catch the subtle mistakes. Philosophical reasoning is especially difficult for AI: unlike programming or mathematics, it depends less on fixed rules and more on conceptual precision and contextual understanding. I don't think there is any doubt that it will help refine our thinking, but I'm not sure that it will replace humans in this area, but who knows.