Thanks for the pointer. Newman and Peirce were saying much the same thing. Peirce developed it more broadly as the mathematical logic – introducing his sign of illation – that then justified his pragmatic approach to truth. — apokrisis

Interesting. I knew Peirce did something similar but with more formality and precision, so this is a helpful lead.

Good. Then we are agreed that abduction, considered as inference to the "best" explanation, does not determine one explanation, and is not itself a rational process. Do we also agree that as a result it doesn't serve to answer Humes Scepticism — Banno

Do you agree that inference to the best explanation can warrant a belief? This of course is only if it was done rationally.

If so, then explain how this doesn't answer Hume's scepticism.

And yes, I am discussing the use of the term, and so its meaning. What I would bring out is that it is not so easy as some might suppose to set out what is a conspiracy theory and what isn't. That the detail is important. — Banno

Are there NO easy cases, in your opinion?

Do you agree that inference to the best explanation can warrant a belief? This of course is only if it was done rationally. — Relativist

Well, I won't disagree, but point out that "the best" remains ill-defined. If we are in agreement as to which explanation is the best, then we should accept it; but here, "the best" might just be "the one we accept"...

Which is to say that we accept the answer based on experience and custom.

So it's not so much answering Hume as agreeing with him.

Are there NO easy cases, in your opinion? — Relativist

:blush: Pretty much. Welcome to philosophical analysis.

Well, I won't disagree, but point out that "the best" remains ill-defined. If we are in agreement as to which explanation is the best, then we should accept it; but here, "the best" might just be "the one we accept"... — Banno

"The best" is always the one we accept. But if the ensuing belief is warranted, that's all that matters. There's isn't a recipe for warrant.

But if the ensuing belief is warranted, that's all that matters. — Relativist

Isn't "warranted" just another way of saying "best"? If it's best, then it's warranted, and if it is the one warranted move, then it's the best?

Further, both involve an evaluation that is not a deduction.

Hume's scepticism — Banno

Following Kant the transcendental subject, through categories (pure concepts of understanding) and their schematism, is the one who imposes the rule of causality, for example, on sensory intuitions. For a sequence of perceptions to become an experience of an object, the "I think" must have unified it according to the necessary rule of cause and effect.

Kant would speak according to the tradition that a belief is true if the judgement formulated by the understanding (through categories) coincides with the sensory intuition that has been previously ordered by the a priori forms of sensibility.

Also Kant rejected the idea that the necessity of natural laws could arise from mere habit. Habit is contingent (it might not exist) and subjective (specific to an individual or species). If science were based on habit, its laws would be mere probabilities, not universal and necessary truths.

So a legitimate belief, according to Kant, is not legitimate if it is based on habit.

That's a very compacted account, yet pretty much spot on. Yep, for us to experience objects and events as structured and intelligible, the transcendental subject applies categories like causality, using the schematism as a bridge, and unifying perceptions under the “I think.”

Is your aim here to understand Hume, and then to move on to Kant's response? that is, are you after an exegesis, or are you looking for broad answers? It seems to me that you have a pretty good grasp of both Hume and Kant.

What I've been doing is more about relating these ideas to more recent work. So we have Hume showing that what "I" have access to is just impressions. He points out that it seems illegitimate to infer the patterns of causation and so on that make up our world. His answer is somewhat unconvincing - habit and custom. This all woke Kant, who used a transcendental argument to claim that minds are predisposed to join these impressions into those patterns.

In Wittgenstein, those “necessary conditions” become grammatical rules rather than transcendental structures: not timeless features of mind, but conventions of our shared linguistic practice. In Davidson, the unity of apperception becomes the principle of interpretive coherence: understanding others and the world depends on fitting utterances and actions into a rational, causal, and linguistic web. And in Feyerabend, even that web becomes plural — our patterns of causation and justification are practices that can vary across scientific paradigms.

How's that?

Isn't "warranted" just another way of saying "best"? — Banno

No. Being warranted means to be rationally justified.
A subjective "best" inference may, or may not, be warranted.

Is your aim here to understand Hume, and then to move on to Kant's response? that is, are you after an exegesis, or are you looking for broad answers? It seems to me that you have a pretty good grasp of both Hume and Kant. — Banno

No. But is an interesting path to follow. My intention is to see how different positions respond to Hume's scepticism.

In Wittgenstein, those “necessary conditions” become grammatical rules rather than transcendental structures: not timeless features of mind, but conventions of our shared linguistic practice. In Davidson, the unity of apperception becomes the principle of interpretive coherence: understanding others and the world depends on fitting utterances and actions into a rational, causal, and linguistic web. And in Feyerabend, even that web becomes plural — our patterns of causation and justification are practices that can vary across scientific paradigms. — Banno

That's very interesting. But you've forgotten to relate it to Hume's scepticism.

No. Being warranted means to be rationally justified. — Relativist

We seem to be circling. Being warranted means to be rationally justified, and something is rationally justified if it is warranted. The best explanations are the ones which are rationally justified, and those are the ones that are warranted, and they are the ones we accept. A subjective best inference may not be warranted, but then it would not be the best inference, and so not justified, and not the best.


Help me out of the loop.

That's very interesting. But you've forgotten to relate it to Hume's scepticism. — JuanZu

Each of those - Kant, Wittgenstein, Feyerabend and Davidson - can be understood as a reply to Hume.

Yes, but choose one of them and explain your answer. I've seen that you like Wittgenstein. So what would Wittgenstein's response to Hume's scepticism be?

How's that? — Banno

All very linguistic. Meanwhile semiotics generalises this to the level of all the codes that are the basis of our interpretative habits. Genetic, neural, linguistic and logical.

Can you explain a little about that in relation to Hume's scepticism? If that's alright.

So what would Wittgenstein's response to Hume's scepticism be? — JuanZu

For Wittgenstein, the demand for justification presupposes a language game, a shared background of practices that already give meaning to “evidence,” “reason,” and “doubt.” To ask, “but how do we really know that causes exist?” is to try to stand outside all language games — to doubt from nowhere. And that, he’d say, is nonsensical. “If you tried to doubt everything, you would not get as far as doubting anything.” So instead of answering Hume’s scepticism, Wittgenstein dissolves it: the idea that we might need a justification for induction or causation presupposes a context of justification that itself depends on certain unquestioned practices.

In a sense it's the same answer as Hume - it's just what we do.

For Davidson, we make sense of the talk of others by presuming that they are rational and are participating in much the same word as we are. We then have the shared basis of the world in order to make sense of the talk of others. Again, we talk in terns of causes in order to make sense to each other.

That's much the same as the basic semiotic answer, that causation and belief are not primarily “out there” in the world; they are enacted and intelligible only within structured systems of interaction, whether psychological, linguistic, or semiotic. Apo and I do not differ as much as might be supposed. What Davidson, Wittgenstein, and semiotics all emphasise is that causation and belief are embedded in structured practices — psychological habits, language games, or broader semiotic systems. They are not features of the world independently of these practices. In that sense, the difference between Apo’s semiotic framing and the Wittgensteinian–Davidsonian account is largely one of emphasis: one highlights codes, the other interpretive and linguistic practices. But both converge on the insight that causal talk is intelligible only within a shared, structured system of interaction.

We could move on to the neuroscience, and talk about how neural nets recognise patterns - a sort of physiological background against which this stuff plays out.

Can you explain a little about that in relation to Hume's scepticism? — JuanZu

Sure. So semiosis speaks to pragmatic reality modelling at four levels. The construction of a model of the world as it is with “us” acting in it.

Even the genes are doing that in terms of making and maintaining a body and the metabolism that sustains it. The immune system, digestive system, hormone system, and all the other body systems are run by intelligence. A model of a self in its world. A set of habits of interpretance.

The Humean issue is framed at the level of science rather than sociology. It is about mathematico-logical semiosis rather than linguistic. Numbers rather than words.

It is not really a sociological issue, or even an neurobiological issue, although all levels of semiosis obviously rely on each other in hierarchical fashion. Genes have to do their job well to sustain a body that has the complexity for neurobiology to then add its level of difference. What use is a clever brain until it is in a body with free hands, for example. Or in a body that really needs extra intelligence to model the social world of a social animal.

So any semiotic relation with the world has some version of causation ain’t correlation. Yet history says evolution has gotten very good at this modelling business. Brains can figure out their environments in practice. If we want to understand the wherefores of epistemology, just see how nature does it.

Humans developed language as a new level of coding and reality modelling. We became socially constructed as members of a tribe, existing in a landscape with an ancestral history. An oral memory and a narrated sense of being selves in world alive with social intricacies.

A new level of semiosis that saw sapiens sweep the Neanderthals aside and colonise the planet. An oral lifestyle that was powered by foraging. One not too concerned with Humean worries about narrations just being narrations. One indeed quite unconcerned as the world seemed animistic - itself part of the collective subjectivity. The trees and the winds all shared our socialised state of mind.

So scepticism is something new that arises when we get to a still higher level of semiosis where the abstract and the concrete creep into the conversation. The idea of objective truth and how it might be secured. The two things we can be certain about because we have a logic to frame the structures that are our general abstractions and a number system to count the concrete particulars or name the material particulars.

Hume voiced a concern that spoke to the transition from a linguistic and sociological level of semiosis to this new rational and scientific level of semiosis. It became important that we were now having to draw a clear line between the subjectivity of our social selves in an oral world and our rational selves in a world demanding objectivity in terms of causal theories and concrete measurements. A strange new world of equations and variables.

Hume became important because he marked this historic crossing point. But once we got used to how this new level of world modelling operated, it soon became a familiar pragmatic habit and Hume’s worries a moment consigned to the history books.

We are now instead at a moment in epistemological theory where we can look back and see that semiosis - this modelling relation - is so general that it is even the basic theory of life and mind itself.

Friston’s Bayesian Brain and free energy principle even dare to put it into a differential equation itself. If your interest truly is in epistemology, it is now testable scientifically theory. Almost to the point of self-caricature.

And if you are still piddling around with Feyerabend, Wittgenstein and Davidson (who he?), then that’s so 1960s. Way behind the times.

The best explanations are the ones which are rationally justified, — Banno

Here's your error. The "best" in an IBE is not necesarily warranted (rationally justified). It just means it was chosen as "best" because it was subjectively judged to be better than alternatives that were considered.

Here's your error. The "best" in an IBE is not necesarily warranted (rationally justified). It just means it was chosen as "best" because it was subjectively judged to be better than alternatives that were considered. — Relativist

So an inference to the best explanation is actually an inference to any explanation?

Maybe I misunderstand, but you seem to be implying that humans have the magical power to select the objectively best explanation from the third realm of abstract objects.

Rather, the "best explanation" has been selected subjectively; the subject has judged it to be the "best" explanation from among the ones he's considered. That is not "any" explanation; it is not arbitrary. But it is subjective, and cannot be otherwise.

The subject may, or may not, have been sufficiently rigorous - he may have overlooked facts; he may have not considered the plausibility of the assumptions he's made or that are entailed; he may have jumped to an unjustified biased conclusion....it's just his judgement.

But surely SOME IBEs, that SOME people make are sufficient to warrant a belief. If not, then nobody has much in the way of warranted beliefs, except for some analytic truths. A corrollary of my claim: a belief can be warranted even if it is possibly false.

So an inference to the best explanation is actually an inference to any explanation? — Banno

It is the one that reveals itself in making quick progress. It has that balance between being causally general and empirically verifiable. One can already see how consequences can be deduced and predictions verified.

Any old guess could be a starting point. The null hypothesis indeed would be the gold standard. There is no actual effect as there is no cause. Whatever you think you are seeing, that is an accident. A spurious correlation.

But then a best guess is any guess coherent enough to get the inquiry started. The one to ride until something better comes along. If the guess Is properly constrained by the null hypothesis, then you can call yourself a scientist.

At issue is finding a solution to Hume's scepticism. That is, how we can move from a finite set of observations to the "objectively best" general conclusion. We know that this is not something that can be done by a valid deduction.

In order to answer Hume's scepticism, abduction would have to show us how to infer the objectively best general conclusion.

We agree that abduction does not do this, but provides only the subjectively best general conclusion.

Hence abduction does not provide and answer to Hume's scepticism, but rather agrees with it.

Any old guess could be a starting point. — apokrisis

We could apply a Bayesian calculus to any old guess, and move towards a better guess, sure. That's one possible solution to Hume's scepticism.

Very much along the lines of Davidson.

So your demand for the “objective and best” inference has just been dropped?

It was never what was being claimed by Peirce in his critique of Humean induction?

At issue is finding a solution to Hume's scepticism. That is, how we can move from a finite set of observations to the "objectively best" general conclusion. We know that this is not something that can be done by a valid deduction. — Banno

All we have as guide is past experience, and what seems to work. Apart from instinct, it's all any animal has. Science (and not just science) is a vast mostly coherent web of belief and understanding that has evolved out of such practices. The "objectively best" general conclusion is merely the one most consistent and coherent with that general web of established beliefs and understandings. Of course we cannot have deductive certainty—that is what Hume's skepticism is about. That seems obvious today, but needed to be pointed out in the age of rationalism.

It could be "done by valid deduction" if there were such things as certain premises. Seems like "much ado about nothing" today.

Valid deductive arguments are contingent on their premises. The conclusion is only "objectively best" if the argument is sound.

Except for arguments whose premises are necessary truths, it is impossible to prove a deductive conclusion is "objectively best". It's an unattainable goal. So why criticize only abductive reasoning for being unable to attain the unattainable?

You could say the deductive conclusion is "objectively best" given the premises, but we could add premises to an abduction that similarly identifies the contingency.

I'm not denying there's a problem of induction:we can't conclude strict impossibility based on a conjunction of evidence. But neither can we conclude a deductive conclusion is a necessary truth, in most cases. This doesn't imply we should all adopt extreme skepticism.

And this is to agree with Hume.

Here's the OP:

In Hume, legitimate beliefs exist. They occur in a process of recurrent association. A belief is legitimate when it is associated with a vivid impression. For example, the belief that one object will move after another is based on past experience of their constant conjunction. Hume concluded that fundamental beliefs, such as the existence of an external world or the existence of the self, are not rationally justifiable but are legitimate because they are the result of experience and custom. — JuanZu

Seems to me that JuanZu is pointing to the prima facie discrepancy between our being confident in a belief based on being "associated with a vivid impression" and a generalisation that is inferred therefrom. I'd understood that as much the same as Popper's basic statements. Hume didn't have a conception of the Duhem-Quine thesis, of course, so took "vivid impressions" as incontrovertible. I don;t see that we could maintain such a thing nowadays.

Hume didn't have a conception of the Duhem-Quine thesis, of course, so took "vivid impressions" as incontrovertible. I don;t see that we could maintain such a thing nowadays. — Banno

And so it continues. Appear to be defending representationalism and then slide away into enactivism. Never stick to an argument so as to slither around and leave a little chorus of lost ducklings in your wake.

Thank goodness AI can clarify what was at stake. Banjo never will...

Charles Sanders Peirce challenged David Hume's skepticism about induction by reconceiving the purpose of inductive reasoning and framing it as a self-correcting process of inquiry. He argued that Hume's critique was based on a flawed, static view of induction, failing to grasp its long-term, statistical nature within a community of scientific investigators.

Hume's problem can be summarized by two points:

Inductive reasoning is not grounded in deductive logic; it is a "probabilistic leap of faith".

We cannot rationally justify the belief that the future will resemble the past.

Peirce's objections to Hume's skepticism include:

The long-run statistical justification of induction: For Peirce, the validity of induction does not rest on guaranteeing a single conclusion. Instead, it is justified by its statistical reliability over the long run. The scientific process, as a collective and ongoing endeavor, employs induction to correct its errors over time.

A "psychological" vs. "real" doubt: Peirce argued that Hume's skepticism relied on a "paper doubt," an artificial pretense of questioning everything. In contrast, Peirce's philosophy of pragmatism asserted that genuine doubt is triggered by genuine surprise or a disruption of a settled habit of belief. Since scientists have no genuine doubt that inductive reasoning produces valuable results, they are justified in continuing to use it.

Rejection of the search for absolute certainty: Peirce accepted that induction does not provide certain knowledge, but he argued that this is not its goal. The purpose of inquiry is not to reach absolute certainty, but rather to fix belief and approach the truth over time. He saw the rationalist search for absolute certainty through deduction as a futile dream.

The distinction between abduction, deduction, and induction: Peirce refined the traditional view of induction by splitting it into two distinct parts: abduction (generating hypotheses) and inductive confirmation (testing those hypotheses). For Peirce, induction is not a standalone method but part of a three-step cycle of scientific reasoning.

In summary, Peirce’s pragmatic perspective reframed the problem of induction. Instead of seeking a rational justification for why a single inductive inference should be trusted, he focused on the practical and statistical reliability of the inductive process as a whole. This process is used by a community of inquirers to progressively refine their beliefs over time, even if absolute certainty remains elusive.

This doesn't imply we should all adopt extreme skepticism. — Relativist

I don't see anyone here suggesting extreme scepticism - including Hume. His point seems to be pretty much the one you are now making.

There is a normative element to deduction, of course, but it is well-bedded, model- theoretical. All that need be accepted is that a predicate is satisfied by the objects of its extension, and that truth is preserved under valid inference. The “normative” aspect here does not consist in a choice among alternatives, but in adherence to what follows from the model itself.

We cannot rationally justify the belief that the future will resemble the past.

We cannot justify it by deductive reasoning, but we can by inductive reasoning—so the conclusion that the future will resemble the past is not certain, but is the IBE. To put it another way, it is rational in a practical sense to assume that the future will resemble the past, because to our knowledge it always has.

Cheers, Apo. Tell me more about me. Yes, the representative vocab can be understood enactively.

Bayes formalises Peirce’s approach if you like. We can move on from the nineteenth century.