Posts Chapter 9 The Architecture of Knowledge

Chapter 9 The Architecture of Knowledge

To what extent does this relation in which one belief is based on another represent our belief system as a whole?

Inferential Chains & The Structure of Belief

Suppose you see the trees off in the landscape swaying. I immediately realize there is wind. The belief that there is wind is on the basis of my belief the trees are swaying. In this particular example the inferential belief, justification, and knowledge is indirect, since the belief, justification, and knowledge is on the basis (or through) other beliefs, justifications, and knowledge. By contrast, the believe the trees are swaying is direct (perception).

Infinite Inferential Chains

A question or observation is perhaps all our beliefs could be indirect. This is an important question and the point here is to explore how people’s beliefs are structured. Let’s start with an example: I’m being cautious, my belief that the trees are swaying could be based on my belief that I have a visual impression of swaying. Could the latter belief be based on another? Might I believe that it seems to me that I have a visual impression of swaying and base my belief that I have an impression on this new belief? Doubtful.

This example illustrates that beliefs are under our direct control of our wills. This is a strong version of doxastic voluntarism (voluntarism about belief) - is a mistake. One simple cannot will a belief. You can’t just accept testimony that seems false as a belief.

Circular Inferential Chains

Imagine holding one belief on the basis of a second and a second on the basis of a third, and so on, until coming back to the first. This would indicate that my beliefs are indirect, but on infinite number isn’t needed. Suppose for the sake of argument, I do have a circular chain of beliefs - there are some problems.

  1. There is good reason to think that (a) one belief is based on a second only if the second is at least in part causally responsible for (for holding) the first. If I believe that there is a wind, on the basis that the trees are swaying, then I believe that there is wind, at least a part because I believe that the trees are swaying.
  2. There is good reason to think that (b) if one thing is in part causally responsible for the second and the second is in part causally responsible for the third, then the first is causally responsible for the third.

Together these points imply (c) in a circular chain of beliefs, every belief is in part causally responsible for itself. That seems impossible. Instead of thinking of each relation being connected with and use supporting this might make more sense. The problem of this self-support is self-justification of an indirect belief doesn’t make much sense. So we conclude there are some direct beliefs.

The Epistemic Regress Problem

We just discussed beliefs previously, but is knowledge the same - some of it is direct or could all knowledge by indirect based on other knowledge we have? It seems possible and that there can be an infinite epistemic regress - an infinite series of knowing each based on the next.

Assume that a belief constituting indirect knowledge is based on knowledge of something else. The further knowledge might be based on knowledge of something still further, and so on. This sequence is called an epistemic chain. It is held that there are four possible kinds of epistemic chains. (1) Infinite-ending unanchored, (2) circular-ending unanchored, (3) ends in a belief that is not knowledge - anchored in sand, and (4) end with a belief that constitutes direct knowledge - anchored solid.

Infinite Epistemic Chains

The biggest difficulty with this case is even if I could have an infinite number of beliefs, how could I ever know anything if knowledge required an infinite chain? Psychologically we just can’t handle them as humans. Even with math we can know 2>1 and 3>2 and so on, but there’s a limit to how far we can go.

Even if it was possible to have an infinite set of beliefs, infinite epistemic chains do not account for all knowledge. For example, the belief that if some dogs are pets, some pets are dogs. There isn’t another belief that this one is based on - it’s self evident.

Circular Epistemic Chains

The circular version has been taken more seriously. It may seem that indirect beliefs cannot be from circular causal chain, so the same is true for knowledge. But maybe knowledge can be based on premises in a way that differs from the way belief is based on them. Perhaps my knowledge that there is wind could be somehow based on my belief that the leaves are swaying, even though my belief that there is wind is not based on my further belief. We could have a circle of knowledge, but not a belief. How realistic is this?

Let’s try it for knowledge. I know there is wind. I know this on the basis of the swaying of the trees. Now I think I know they are swaying because I see them sway. It could be argued that my seeing is only on the causal basis of my belief that they are swaying. It seems hard to go full circle unless I think up propositions I do not originally believe (do not know).

Even if I could come up with a suitable set of propositions. Suppose the visual impression of swaying. Might I know this on the basis of knowing there is wind?

Epistemic Chains Terminating in Belief not Constituting Knowledge

This particular type is best understand on source conditions and transmission conditions. If one transmits when there is no justification then one ends up with no knowledge. If it were foggy, you couldn’t see the trees swaying, but guessing correctly - you would believe there is wind. Knowledge can’t be grounded in guess work.

There is the possibility of making an educated guess - considering more evidence - to reach the conclusion. It is still a guess and there for not grounded knowledge, but on the degree of evidence considered could be close.

Epistemic Chains Terminating in Knowledge

This is the favored possibility - epistemic chains end in direct knowledge. Normally I know that there is a swaying just by seeing that there is. The chain grounding my knowledge that there is wind is anchored in my perception.

These experientially or rationally grounded epistemic chains may differ in many ways:

  1. Composition: Differ in composition in the sorts of beliefs constituting them.
  2. Transmissions: Differ in the kind of transmission they exhibit - deductive, inductive, or both.
  3. Anchors: Differ in their ultimate grounds - experiential or rational.
  4. Strength: Differ in justification strength - the degree of justification they gave to the initial belief.

The Epistemic Regress Argument

What has been discussed suggests a version of the epistemic regress argument.

  1. If one has knowledge, it occurs in an epistemic chain. Epistemic chains are understood to include a special case of a single link - such as perception.
  2. The only possible kinds of epistemic chain are the four previously discussed: infinite, circularly terminating in beliefs that are not knowledge, and those terminating in direct knowledge.
  3. Knowledge can only occur in the fourth kind of chain.
  4. If one has any knowledge, one has some direct knowledge.

Basically it is understood as such (1) implies that any given instance of indirect knowledge depends on at least one epistemic chain for status as knowledge. Simply put any indirect knowledge requires an epistemic chain to some direct knowledge held by the person. A similar argument can be made for justification.

Foundationalism & Coherentism

Epistemological foundationalism - the structure of the body of knowledge is foundational, which is taken to imply that any indirect (hence non-foundational) knowledge there is depends on direct knowledge. The structure of the body of justified beliefs is foundational, where this is taken to imply that way indirectly (hence non-foundational) justified beliefs there are depend on directly justified beliefs.

A critique of foundationalism is coherentism. The central idea is that the justification (justifiedness) of a belief depends on its coherence with other beliefs one holds. The units of coherence (range of beliefs that must cohere in order for a belief among them to derive justification from their coherence) may be as large as one’s entire set of beliefs.

Holistic Coherentism

Coherentism need not be linear - it can be holistic. An example is the best way to illustrate the holistic approach. John wonders how I know, as I sit reading, that the wind is blowing. I say that the leaves are rustling. He asks how I know that Sally is not just making the noise by walking in high grass. I reply that the high grass is too far away. He wonders if I can distinguish rustling leaves from the sound of a quiet car on a pebbled driveway. I reply that what I hear is too much like a whisper to be the crunchy sounds of a car on pebbles.


In this example, there is only one step taken on the inferential line. The original belief (the wind) are defended in terms of the entire pattern of mutually cohering beliefs I hold. On the coherentist view, beliefs representing knowledge do not have to lie in a grounded chain; they fit a coherent pattern and their justification emerges from their fitting that pattern in an appropriate way.

Coherentist Response to the Regress Argument

The main point that can be said is a denial of there only being four kinds of epistemic chains. The coherentist would contend there is a fifth kind of epistemic chains. A chain terminating with the belief that is psychologically direct, yet epistemically indirect (or in the realm of justification, justification ally indirect.) This point grants foundationalists that they’re right about human psychology, while insisting they are wrong about epistemology.

Direct belief is not psychologically based on any other, as when it is inferentially grounded on another, its justification nonetheless is based on other beliefs. Hence, the last link is, as belief, direct, yet, as knowledge, indirect in the broad sense that the belief constitutes knowledge only by virtue of receiving support from other knowledge or belief. This belief is psychological foundational but epistemically dependent.

The Nature of Coherence

It’s somewhat difficult to explain what coherence is. It’s not the same sort of mutual consistency, but inconsistent is the clearest case of incoherence. Two propositions that are completely unrelated, like 7+5=12 and carrots are nutritious are mutually consistent, but not coherence.

Coherence and Explanation

The idea of coherence is that it is sometimes connected with explanation. So belief in a second propositions explains a belief in the first proposition. The process of explaining something is that makes it understandable. Making understandable is a coherence - generating relation between propositions.

The concept of probability is also relevant to coherence. The probability of a proposition at least counts in favor of my belief of the first cohering with my belief of the second. How does explanation and probability understand with coherence?

I can suppose a genie delivered an unmarked package to me. It does explain the delivery of it. I wouldn’t be justified in believing it. It is difficult to say when an explanation is relevant to understanding coherence. It’s difficult to specify when an explanatory relation generates enough coherence to create justification.

Similar points hold for probability. Suppose you had a second proposition you didn’t believe at all, but was a great way of explaining the first. It would raise the probability of believing the first, though, I’m not thereby entitled to believe it.

Coherence as an Internal Relation Among Cognitions

So far as we understand coherence, what reason is there to think that by itself it generates any justification or counts toward truth?

The isolation problem is the problem of explaining why coherent systems of beliefs are not readily isolated from truth, and thus do not contain knowledge, which implies truth. Also the problem of explaining why there is not a similar isolation from justification, which seems to point toward truth - roughly what justifies a belief indicates truth - and indicates it in proportion to the degree of justification.

Why should coherence by itself imply that any of the cohering beliefs is justified or constitutes knowledge, when both justification and knowledge point toward truth as something external to the belief system?

Let’s suppose that I think I’m Napoleon. I could very well have a whole belief system to explain and justify my belief as Napoleon. My therapist may have a totally different coherent beliefs on me. If Napoleon’s and the therapist’s belief systems are equally coherent, how can we justify our apparently quite reasonable tendency to regard their belief systems as more likely to represent truths?

We might take the therapist’s belief that I’m not Napoleon, since it’s their field, but why should our own beliefs be privileged over equally coherent conflicting sets?

Coherence, Reason, and Experience

Another problem with coherence is how to explain the role of experience and reason as apparent sources of justification and knowledge.

A coherentist may say that many of our beliefs are causally and non-inferentially based on perception or reason; and given these similarities of origin, it is unsurprising that they often cohere with one another. Basically what we perceive/reason need not infer propositions, they do cohere with many things we believe and this coherence is what justifies them.

Coherence and the A Priori

If we take the proposition: some dogs are pets then some pets are dogs, can be justified for me even if it does not explain or justify anything else I believe. So why is coherence required for my justification.

A coherentist might say that you apply this sort of view to empirical beliefs, rather than a priori propositions. This defended on the assumption that propositions known a priori are necessarily true, and therefore are not appropriately said to be made probable by other propositions.

This isn’t the best reasoning or the coherentist point. Imagine that someone mistakenly takes a certain false proposition to be a theorem of logic and cannot see why a closely similar, true proposition is a theorem.

Coherence and the Mutually Explanatory

Mutually explanatory: no proposition is both true and false. It’s truth explains why they hold, and their truth explains why it holds; and this is the chief basis of their mutual coherence.

Reflection on other purported examples of mutual explanation also suggests that true propositions cannot explain each other. A man could say something because his wife did, and she said it because he did. Now one of them would have to say it first to cause the other. Suppose we attach a time constraint. He said it because she said it earlier. Let’s take it further. He said it because she said it earlier, and she said it because he said it earlier. This is the kind of reciprocal explanation, wherein a kind of thing explains and is explained by another thing of the same kind. But this is not a mutual explanation, wherein the very same thing explains and is explained by the second thing.

If coherentism applies only to empirical beliefs and not to beliefs of a priori propositions, then its not a general theory of justification or knowledge and leaves us in need of a non-coherentist account of a priori justification.

Epistemological versus Conceptual Coherentism

Foundationalism is in fact consistent with one kind of coherentism, a coherence theory of the acquisition, function and nature of concepts. Put short, the coherence theory of concepts - on which concepts are what they are partly in relation to one another, and a person acquires concepts, say of physical objects and shapes, and of music and sounds only in relation to one another and must acquire an entire set of related concepts in order to acquire any concepts. Meaning the concept of music includes the concept of sound. Or a piece of paper includes the concept of rectangle.

If the coherence theory of concepts is sound, a foundationalist must explain how it squares with their epistemology. The control point they may appeal to is a distinction between grounding conditions for belief and possession conditions for it. What grounds a belief in such a way as to justify it or render it an item of knowledge is largely independent of what other beliefs one must have, and what concepts one must have, to be able to hold the first belief.

The point here is simply that we cannot treat conditions for having a belief at all as doing the more specific job of grounding its justification. By and large beliefs can be possessed without being justified , and their is usually a good distance between meeting the conditions for simply having a belief and meeting the standards for justification is holding it.

Positive and Negative Epistemic Dependence

It is essential to distinguish negative epistemic dependence - which is a form of defeasibility - from positive epistemic dependence - the kind beliefs bear to the source from which they derive any justification they have, or if they represent knowledge, derive their status as knowledge.

The defeasibility of a belief’s justification by incoherence does not imply that this justification positively depends on coherence. If a well is a source of water, I’m positively dependent on it. It is possible people could drain it, doesn’t imply a (positive) dependence on them, such as I have on the rainfall. It is the rainfall that explains both my having water and it’s level - it not being drained does not explain it.

Coherence and Second-Order Justification

Often, in defending a belief, one forms new beliefs, such as the belief one acquires in moving one’s head that one can vividly see the changes in perceptive that go with seeing a bird on a branch.

The Process versus the Property of Justification

These new backup beliefs are appropriate to the process of justifying one’s belief - the result of this process is showing that the original belief is justified, together with one farming a certain second-order belief - because it is a belief about a belief, which is not about another belief.

Beliefs, Dispositions to Believe, and Grounds of Belief

I say that there is a green field before me. I can give my justification: that I see it. First, giving a justification is not equivalent to claiming that one has it. The first cites a justifier and need not employ the concept of justification. Second, before justification I need not even believe that I see a field. Basically, one has a belief of a field, not the perceptual relation to it.

A disposition to believe something does not imply one’s already having a dispositional belief of it. So basically my belief about the field is not by appeal of coherence with other beliefs I already hold, but by reference to a basic source - sensory experience.

Justification, Knowledge, and Artificially Created Coherence

If coherentism regards justification as derived from coherence alone, then it accords no justificatory weight to experiential or rational grounding - except insofar as they contribute to coherence.

The problem here is that I could have a superbly coherent system of beliefs I might acquire that run counter to my experience. A coherentist might reply that if we are talking not only about justification, but also about knowledge- then we must give some special role to beliefs grounded in experience and reason.

Moderate Foundationalism

So far we have seen there are for more problems for coherentism and foundationalism. The most serious problems foundationalism faces:

  1. The difficulties of specifying source conditions for justifications and knowledge.
  2. Of accounting, on the basis of those sources and plausible transmission principles, for all that we seem to know.

The first point was addressed in the first part of the book. The second is treated in chapter 8, which indicates how knowledge and justification can be transmitted from beliefs which are justified by virtue of being grounded in the basic sources, to other beliefs. Both problems are difficult, and neither has been completely solved.

The Role of Coherence in Moderate Foundationalism

Another problem for foundationalism is the difficulty of accounting for the place of coherence in justification. Foundationalism can account for some of the insights of coherentism. More positively foundationalism can acknowledge a significant role for coherence in relation to justification and can thereby answer one traditional coherent objection. A kind of moderate foundationalism: a foundationalist view of knowledge or justification which (1) takes the justification of foundational beliefs to be at least typically defeasible; (2) is not deductivist, that is, does not demand the principles governing the inferential transmission of knowledge or justification be deductive (require entailment as opposed to probability as a condition for transmission), not that inferentially justified beliefs derive all their justification from foundational ones, but only that they derive enough of it from that latter to remain justified if any other justification they have were eliminated. Some versions of moderate foundationalism differ, but the most plausible ones give coherence at least two roles.

First role - is negative: incoherence may defeat justification or knowledge, even if directly justified (foundational) belief, as in I may be hallucinating. Second role - an independence principle that the larger the number of independent mutually consistent factors one believe (with some justification) to support (or constitute evidence for) the truth of a proposition, the better one’s justification for believing it.

Moderate Foundationalism and the Charge of Dogmatism

Moderate foundationalism avoids dogmatism. Epistemological dogmatism - the version to us indefeasible justification, epistemic certainty, where these attributions are unwarranted by our violence.

All Rights Reserved.