Knowledge and Justified Beliefs

1. Skeptical Arguments

In the ordinary everyday sense of the word, a skeptical person possesses high standards for what they will believe in, and what they will consider knowledge. They are not easily persuaded. The beliefs they hold undergo rigorous verification standards. But they do not doubt the possibility of knowledge— they merely find the epistemic standards of most people too unrigorous. Skepticism in a philosophical sense goes further in two aspects:

  • Scope: Philosophical skepticism questions even the possibility of knowledge. It claims knowledge, in general or in certain areas, is impossible, and thus "aspiring to knowledge - either everywhere or with respect to certain very broad domains of fact - is inherently problematic."1 This gives this skepticism a very broad, perhaps all-encompassing, scope.
  • Degree: There are also variants of philosophical skepticism which are radical. They not only question whether knowledge is possible, but even whether beliefs can ever be justified. This is seen in the distinction between philosophical skepticism and fallibilism. The former thinks we do not even have reason to believe one thing or another, in other words, that we are never justified in believing anything. Fallibilism claims that beliefs can be justified, that we can have a good reason to believe something, but that justification does not entail knowledge.

It is important to solve the problem of philosophical skepticism because knowledge is relevant to anyone who makes use of it, namely everyone. The skeptical conclusion that knowledge is impossible is unnacceptable because we constantly depend on knowledge in order to live. And though knowledge may not exhaust the realm of important propositions which we must consider (others may include moral propositions for example), we could not live at all if we took seriously the claim that no knowledge or justified beliefs exists. For this reason, 'in order to live', no one takes it seriously, that is, we simply assume that knowledge is possible. Does this assumption make this discussion a moot point? If we're not going to accept skepticism anyways, why bother debunking its arguments? Williams (2017) argues that even if we already reject their conclusion, skeptical arguments are still worth analyzing and debunking for methodological reasons. Firstly, they are very good arguments. They are logically airtight and have such a strong intuitive appeal that we cannot just dismiss them, at risk of being dogmatic and philosophically unrigorous. Furthermore, dissecting good arguments and potentially refuting them could teach us a lot about knowledge. There exist two broad families of skeptical arguments: the variants of Agrippa's Trilemma, and the variants of Cartesian skepticism. In the next section, I will briefly summarize their arguments, after which I will go on to survey various attempts at refuting them.

Agrippa’s Trilemma

Agrippa's trilemma goes as follows: Say you claim something is true. It is perfectly reasonable to ask you how you know it to be true. You might then answer with some justifier, for example some empirical evidence, or some principle you believe to be rational. But then, this justifier is itself a claim that you assume to be true, otherwise it could not ground your first claim. So again, the question is asked how you know the justifier to be true. So to justify the justifier, you invoke another claim. Again, it is reasonable to ask how you know this second justifier to be true, and so on it goes. At each renewed demand for justification, there are three possible ways for you to respond:

  1. You give another justifier and continue this process infinitely. This is called Infinite Regress.
  2. At some point, you stop at a given claim without further justification. The foundation of your chain of justifications is thus a dogmatic assumption. This is called Foundationalism.
  3. You repeat a justification you made earlier, thus making your justification chain circular. This is called Coherentism.

Agrippa's Trilemma adds the premise that none of these options are satisfactory, and so concludes that no valid form of justification exists whatsoever.

As an argument, it is deeply intuitive because its main move is to demand justification for each and every claim in the chain of justification. We intuitively feel that it is always valid to ask how someone knows some claim is true, that this is a valid response to any and all claims to knowledge. There seems to be a universal assumption that "knowing differs from merely assuming or surmising and that this difference has something to do with an ability to back up or justify whatever can properly be said to be known."2 People feel that if they cannot answer the question of how they know something to be true, then they do not truly know it. Furthermore, the fact that it is a Trilemma makes it very adaptable, capable of arguing against a wide diversity of non-skeptical positions. Depending on the non-skeptical position, one of the three prongs may pose a greater problem for it than others. As such, the trilemma is an argument rather than a specific problem; an argument that can be employed against many different problems and stances.

The conclusion of the Trilemma is deeply problematic for non-skeptics, both because of its degree of skepticism and scope. Regarding degree, it results in radical skepticism because it concludes that not even (valid) justification exists, let alone knowledge.3 Its scope is general, i.e. its skeptical conclusion applies to all knowledge, because it does not only apply in conversations between two or more people where someone actually demands justification for claims. It applies to all individuals even when they are alone with their thoughts. Whatever the thought or claim, we feel it is valid to ask ourselves: How do I know this is true? So the Trilemma applies to all our beliefs, all our claims, all possible propositional content. It doesn't get more general than that.

Cartesian Skepticism

The second family of skeptical arguments was first fully formulated by Descartes. It concerns our beliefs and explanations about our sense experiences and the external world. Our experiences, what we feel, perceive, and remember, are such as they are. Being as they are, "non-propositional and non-conceptual",4 they are 'given', undistorted, assumptionless, and require no further justification in themselves (that is, if we do not try to say anything about them). But our descriptions and explanations of these experiences are not self-evident. Descriptions and explanations are propositional in content structure, and there are infinitely many possible descriptions, explanations and systems of belief that fit equally well with our non-propositional experiences. So why are we so sure that our current beliefs and explanations are the true ones, among these infinitely many options that fit equally well with our experiences? This question, this doubt, makes Descartes skeptical about whether knowledge of the external world is possible.5

He argues for this skepticism by illustration. Throughout life, we experience such things as optimal illusions, hallucinations, and dreams. The consensus among us is that dreams are not 'real'— they do not reflect reality and are not to be trusted as sources of truth. But in principle, Descartes says, it is possible for dreams to vividly mimic our moments awake, the 'real' moments, completely and faithfully. So why are we so sure the entire thing is not a dream? Our time awake could also be a dream, perhaps with experiential features that are often more time and space consistent than our dreams while sleeping. Descartes' second, more famous example, is that some omnipotent demon is deceiving you (the conscious subject) by manipulating you, so that you experience all of that which you experience: the physical world, objects, other people etc. But all of this is in fact an illusion, and the real world is completely different. The modern version of this is that you are just a brain-in-a-vat, where aliens manipulate you to have experiences identical to a brain in a body. The point with all these scenarios is that they are plausible, coherent alternative explanations of our experience that fit equally well with our

experiences. So how can we know that we are correct in our current explanation? Descartes argues we cannot.6

Descartes' skeptical argument gave rise to many other variants not invented by Descartes himself. They make use of the same argument, that there exist many plausible alternative explanations for phenomena, and that one therefore cannot know which is true. Examples of these variants of Cartesian skepticism include:

  • Solipsism: The view that I can only know that I exist, but not that others exist. I only experience my own experiences, not others'. I might try to imagine the experiences of others, but I can never actually have experiences other than my own. Consequently, I cannot truly understand the experiences of others, nor can I use the same terms for psychological states, because who knows if I feel happy in the same way you do, or if you feel happy at all? More than this, I cannot even know whether any conscious subjects other than myself exist, since experience is the only thing 'known', and I only experience my own experiences.
  • Skepticism towards the Past: We experience certain memories and empirically perceive certain historical artefacts. But we cannot know the explanation behind them. We cannot know whether or not our memories were merely implanted by an evil demon, who deceived us into believing we have lived for years, when in fact we have only existed for minutes. And we cannot know whether or not historical artefacts were implanted by some deceptive God.7 Hence, we cannot know anything about the past. We only know that we experience a past.
  • Skepticism towards the Future, or the problem of Induction, initially formulated by Hume: Just because something repeatedly happened in the past, does not imply that it necessarily will continue to repeat in the future. The continuation of empirical patterns is not a necessary truth. That past patterns will continue in the future is an assumption, not a necessary truth.8

This idea was echoed later on, in Philosophy of Science, as one of the main criticisms of logical empiricism. Propositions do not follow directly from empirical observations. True propositions are not served to us on a silver platter as a matter of course following experience. Experiences do not 'imply' certain propositions, because translation of raw experience to propositions is not at all a straightforward process. Such acts of translation are not assumptionless i.e. theory neutral. Cognizing subjects actively participate in the translation process. Perceptual beliefs are thus partially the result of subjective contribution, and not solely a 'reception of objective propositions'. A simple everyday example is the fact that various people at the same scene can interpret the events in completely different ways. A scientific example is that in recording observations, scientists cannot capture the complete sensory 'picture', and are forced to choose certain aspects to focus on, i.e. variables, and also how to measure and record them. All such choices of what to focus on, and also how to focus on them require theoretical assumptions, none of which were self-evident or implied by the observations themselves, and all of which could be questioned and replaced with alternatives that might equally fit with the observations. All this makes observation (and science) more like an art rather than a science, in the way science is commonly understood. Feyerabend (1999), Longino (1990) and the Duhem-Quine thesis discussed in Stanford (2023) are examples of discussions about the theory-ladenness of observation.

2. Direct Responses to Skepticism

Now that we have surveyed the main skeptical arguments, let us look at the various attempts at refuting it. Williams (2017) distinguishes between two types of approaches to refuting skepticism: direct and diagnostic approaches. Direct approaches try to attack skeptical arguments as they are framed by skeptics. That is, they try to refute either one or more of the premises, or to show that the conclusion does not follow from the premises. Diagnostic approaches, on the other hand, do not accept skeptical arguments as they are presented by the skeptics, but claim that the entire way the skeptical arguments are framed is misleading. In this way, they do not respond directly to skeptical arguments, but rather sidestep them. Diagnostic approaches can be further divided into two broad subgroups. The therapeutic diagnosis approach tries to show that skeptical problems are pseudo-problems, i.e. merely apparent and not genuine problems. This can for example be problems caused by incompetent and confusing use of language. An example of such a view is the internalist contextualist account proposed by Cohen (1999). Then there is the theoretical diagnosis approach, which analyzes skeptical arguments to reveal their implicit assumptions. These assumptions are then argued to be objectionable. The externalist contextualist account by Williams (2017) is an example of an account which takes this approach.

Direct responses to the Agrippan Trilemma accept the premises that indeed, it is always epistemically valid to ask someone to justify their beliefs, and that indeed, a person being asked for justification can only reply in one of the three aforementioned ways— towards infinite regress, foundationalism or coherentism. These responses then try to show that one of these three outcomes are epistemically unproblematic and can produce adequate justification. There was quick consensus in the philosophical community that infinite regress was not a viable path, so only coherentism and foundationalism remained as options to pursue. So far, coherentist views have not succeeded in dealing with the objection that coherence is indeed a necessary condition, but not a sufficient condition for knowledge. A most important argument for this objection is that there exist infinitely many coherent systems, but only one is true. How can we know that the one we believe is the true one, when they all equally satisfy the criterion of coherence? So, coherence alone does not entail truth. A coherent system needs to be 'anchored' to reality and truth somehow, by additional required features. This 'anchor' is what enables 'discriminating those that can reasonably be trusted from those that cannot'.9

Foundationalism and its Problems

Foundationalist theories claim that the chain of justification can validly stop at certain foundational beliefs, which themselves do not require further justification— they are 'intrinsically justified'. Thus, the chain of regress comes to an end at these basic beliefs. Differing foundationalist accounts differ in what they claim intrinsic justification consists in, and which things possess such intrinsic justification, i.e. which things are basic beliefs. Many foundationalists have tried to make sensory experience this foundation. But the raw sensations themselves cannot be the foundation of our chain of beliefs because the sensations are non-propositional, while beliefs are propositional. This difference in content structure means that raw experience "cannot stand in logical relations to beliefs and so can nether support nor falsify beliefs."10

Other attempts have then made perceptual beliefs the foundation. We do indeed possess beliefs that arose from sensory perception. But acquiring such perceptual beliefs requires a sort of translation of experience's non-propositional, non-conceptual content into propositional descriptions of these experiences. And "where there is description, there is the possibility of misdescription".11 In reality, most phenomena are so complex that we easily draw the wrong interpretations. Nor is observation itself entire assumptionless and theory neutral.12 Since some perceptual beliefs are accurate and some are not, we cannot say that perceptual beliefs as an entire class are infallibly accurate. The possibility of error makes this class vulnerable to be defeated by the deeply held intuition that misdescriptions i.e. false perceptual beliefs, are not justified beliefs. We thus cannot call perceptual beliefs, as an entire class, intrinsically justified. This makes perceptual beliefs also unsuitable to constitute the foundation of our chain of justification.

Descartes tried to make a necessary truth his basic belief: "I think, therefore I am." Sosa’s (Sosa & Stroud 1994) objection to this was that the necessary truth was not Descartes's cogito, but rather the conditional “if I am thinking that I think, then I am right”.13 But while this conditional is true, nothing makes it necessarily true that everyone would think that they are thinking. That is, it is possible for people to mistakenly think that they are not thinking. "[Descartes] explains incontestably why one must be right in thinking that one thinks, he does little or nothing to explain why it is that the cogito and other similarly simple, clear, distinct propositions are for us indubitable."14 So what Descartes needs in addition to the conditional, Sosa claims, is to show that it is impossible for anyone to mistakenly think that they are not thinking.15

This is a problem that Descarte's cogito foundationalism shares with experience foundationalism: Just because something is obviously or necessarily true (e.g. sensations and logical truths respectively), does not mean that it is indubitable to the cognizer. Experiences can be misinterpreted, and necessary truths can be disbelieved. The obviousness or necessary truth of a belief does not guarantee that cognizers will believe it. This is why, Sosa argues, what matters for justification must be something that does not require the assent of the cognizer, something for example like the actual reliability mechanisms for acquiring beliefs, regardless of whether this reliability is known by the cognizer. ”What matters is not that one attend to the contents of one's mind, to one's experiences or beliefs or other states of mind, nor is what matters that one attend to simple necessary truths (...) It is important rather that the subject be reliable on the object of knowledge, and unlikely to go wrong on such subject matter.”16 This reliable process theory is an externalist account of justification which we shall discuss in more detail later.

Direct Responses to Cartesian Skepticism

All variants of Cartesian skepticism share the following common argument structure:17
P1. There is a gap between our experiences and our 'knowledge' (beliefs) about the world. That is, true propositions do not follow directly from raw experience.
P2. This gap cannot be bridged deductively.
P3. This gap cannot be bridged inductively.
Conclusion: There is no way to bridge the gap = no way to attain knowledge about our experiences. We can never know more than the experiencing of our experiences. We can never have propositional knowledge.

This structure delimits the space of possible direct responses to Cartesian skepticism. Direct responses have tried to show one or more of the premises to be false. Attempts at refuting the first premise have failed, firstly because of Descartes' argument that appearances do not necessarily reflect reality (remember the scenarios of the Evil Demon and brains-in-vats). But they also fail due to the problems of translating non-propositional experience to propositional beliefs extensively discussed in the previous section on foundationalism. Critical Theories have tried to refute the second premise and show that the gap can be bridged deductively, by arguing that there exists an a priori-discoverable relationship of a kind that justifiedly bridges the gap between experience and reality, e.g. an intrinsic, meaningful, or necessary relation. But they also seemingly fail, because "skeptical thought experiments seem to show that there is no particular a priori connection between appearances and reality, even of this weaker sort."18

Attempts to prove that the gap can be bridged inductively have been made by invoking the principle of Inference to Best Explanation. That is, in answer to Descartes' question, such positions take explanatory power to be the proper criterion for evaluating between competing alternative hypotheses. The winning theory is the one that can best account for features characteristic of our experience. We see, then, that claims as to which features are characteristic of experience are crucial in discriminating between alternative hypotheses, ranking them as more or less fitting to experience. Proponents of our current understanding of the world-- i.e. as physical, with physical objects and laws-- argue that this hypothesis fits better than alternative skeptical hypotheses with the characteristic features of our experiences, important ones of which are regularity, stability, coherence, constancy and predictability. Their conclusion that the 'Physics Hypothesis' is the best explanation depends on the premise that these aforementioned features truly are characteristic of experience.

This premise is, however, questionable. It is not obvious that experience is characteristically regular. While experience does undeniably display certain regularities, it undeniably also displays many irregularities. Perhaps we already have unconscious presuppositions which make us notice regularities more often than irregularities, assumptions which smoothen out and make us overlook a lot of the irregularities and incoherence present in 'raw' experiences. Nor is it obvious that experience is stable. Does the same object look the same to every person? What about colors? The answer to this is not obvious even to the community of non-blind people. Are events experienced in exactly the same way despite a change in, say, mood or first-person (perceiver)? It seems highly doubtful. These examples show the great variability and instability of experience. Constancy is not necessarily a characteristic feature of experience-- it is assumed to be so. This problem was first formulated in Hume's problem of induction, where he shows that it is not a necessary truth that past patterns continue into the future. The non-necessity of constancy thereby also defeats predictability.

Even supposing these aforementioned features were characteristic of experience, it is not self-evident that an external physical world should be the best explanation for them. Furthermore, there is a problem with experience itself, with taking it to be the target of explanation i.e. explanandum. Our experience, which we try to infer the best explanation for, is based on our memory. As we have seen, Cartesian skepticism about the past shows that memory does not necessarily reflect the reality (e.g. you have only existed for minutes, but the aliens manipulated your brain-in-a-vat to believe you've existed embodied and for years).

3. The Assumption of Epistemological Realism

For all the reasons discussed in the previous section, direct responses to skepticism tend to fail. Let us then consider some diagnostic responses to skepticism. Williams (2017) proposes a theoretical diagnosis of skepticism. He argues that both Agrippan and Cartesian skepticism are based on assumptions which are themselves questionable.

The Prior Grounding Requirement

There are two dimensions of justification which must be satisfied for justification simpliciter, or knowledge, to be satisfied. One is personal justification, which is attained by being epistemically responsible. The other is evidential justification, which is attained through a belief having adequate grounds or adequate evidence to support it. The central move in Agrippa's Trilemma's is to demand evidential justification for each and every belief in the chain of justification. This implicitly assumes that it is epistemically justified to demand justification at every turn. This assumption is supported by the very deep and widely held intuition "that epistemically responsible believing is always and everywhere believing on the basis of adequate evidence",19 that "belief is epistemically irresponsible unless it rests on adequate grounds."20 Notice that this is the view that personal justification requires, and is thereby subordinated to, evidential justification. If and only if based on adequate grounds or evidence are beliefs personally justified (evidential justification -> personal justification). Williams calls this the Prior Grounding Requirement (PGR), the assumption that the relation between personal and evidential is one where the latter is prioritized.

This assumption achieves two things for the Trilemma Argument: 1) It justifies the demand for justification at every turn, and 2) it allows the Trilemma to generalize the inexistence of evidential justification to the inexistence of all epistemic justification, as follows. The Trilemma demands evidential justification at every turn. This constant demand for evidential justification leads to the three aforementioned prongs. From this it concludes that no valid evidential justification exists. But since, according to PGR, evidential justification implies personal justification, and their conjunction then implies justification simpliciter, the Trilemma can conclude from the inexistence of evidential justification that personal justification and justification simpliciter do not exist either--- the latter meaning that justification does not exist whatsoever.

Epistemological Realism

The reason that the Agrippan Trilemma presupposes the PGR, that it is always justified to demand justification for beliefs, universally, in all contexts, is because it is committed to epistemological realism. This is a belief in an 'extraordinary metaphysics of knowledge',21 the belief that there exists an objective metaphysical epistemic reality. What sorts of epistemic entities this reality is populated with may differ depending on the particular view. But most often, different epistemic kinds are posited, each with an intrinsic epistemic status which may be unequal, thereby creating an objective hierarchy of epistemic status. Proponents of epistemological realism would thus believe in universal and timeless epistemic principles rather than context-sensitive ones. But they can differ in what they take to be these universal and eternal epistemic principles (or facts, or kinds, or relations etc). The Agrippan Trilemma takes the Prior Grounding Requirement to be the universal and timeless criterion of epistemic justification.

Cartesian skepticism also presupposes epistemological realism. Their epistemic 'ontology', their view of what the objective epistemic reality is, consists of two distinct epistemological kinds: internal and external knowledge. They claim that these are the 2 epistemic kinds that exist, and that they are distinct. Internal knowledge is all knowledge gained in our minds, which includes all our experiences and nothing beyond. External knowledge encompasses all knowledge which is 'without the mind' e.g. propositions about the world, the body, god... Cartesians also believe in that experience, i.e. internal knowledge, is epistemically prior and superordinate to knowledge about the world, i.e. external knowledge. This epistemic priority means that if we are to have external knowledge at all, then such knowledge must be depend on or be derived from internal knowledge in some way. In other words, Cartesians' commitment to the epistemic priority of internal knowledge over external knowledge forces them to commit to foundationalism of experience.

It is because of their implicit commitment to foundationalism of experience, that they argued as they did. The Cartesian argument, again, was that there are infinitely many coherent beliefs systems that fit our experiences equally well, but no way to discriminate between them. We lack a valid way to discriminate between these systems, and therefore cannot have knowledge. But notice, the hypotheses that qualified as viable candidates (of truth) consisted only of those which fitted well to experiences. It is because they presuppose that knowledge (true hypotheses) can only be based on experience that they argue thus. So the structure of the Cartesian argument can be represented as we did earlier in this essay, just adding the presupposition of Experience Foundationalism at the beginning of the argument.

That is, the argument can be represented as:
P1: Presupposition: Experiences are the only possible basis of knowledge i.e. foundationalism of experience.
Cartesian Skepticism Argument as before, i.e. that the gap between experiences and propositions cannot be bridged, shows that
P2. Experiences cannot yield knowledge (i.e. foundationalism on experience cannot yield knowledge)
Conclusion: No basis of knowledge exists, so knowledge is impossible.

Foundationalism of experience is a premise of the Cartesian argument leading to skepticism. It is a cause of Cartesian skepticism. I emphasize this point, because many attempts have been made in the epistemological tradition to resolve Cartesian skepticism with foundationalism of experience, not realizing that the latter in fact contributes to causing it.

The Contextualist objection to Epistemological Realism

To summarize, both Agrippan skepticism and Cartesian skepticism presuppose epistemological realism, the existence of certain universal and timeless epistemic entities (kinds, facts, standards, principles, relations or other). They stand and fall with it. Contextualism is the view that epistemic entities are context-sensitive. So for example, a contextualist account of epistemic justification would claim that the standards for being epistemically justified depend on the context. The standards may differ with regard to which criteria must be met, but they may also differ with regard to the degree to which these criteria must be met. This view is obviously incompatible with epistemological realism-- context-sensitive entities are neither universal nor timeless. To Agrippan and Cartesian skeptics, the contextualist would say: Just because the conclusion is skeptical in a context which begins with skeptical assumptions, does not mean that the conclusion is also skeptical in a context with different assumptions. A conclusion that follows from certain premises only holds in contexts that share those premises, and precisely those premises. To take the conclusion from one context and extrapolate it to other contexts with different assumptions is a logical “fallacy akin to equivocation".22

Williams (2017) and Cohen (1999) concede that in skeptical contexts, beginning with skeptical assumptions as presented by the skeptical arguments, the conclusion is indeed inevitably skeptical. In this sense, they are both skeptics. In such contexts, we cannot know that we are not brains-in-vats. But to then extrapolate this conclusion to the context of, say, a physical experiment, and instead of raising a doubt about whether the measurement equipment is adequate, you raise doubt about whether it exists at all, whether objects being measured exist at all, whether you have a body at all, you are changing the subject from physics to (skeptical) epistemology. Such a change in subject is actually irrational and is neither good epistemic nor good dialectical protocol.

Now, if the skeptics claims that their epistemic assumptions hold in all contexts, so that a change in context would not change the conclusion, then this is a claim they must defend. They are not per default justified in taking for granted that skeptical assumptions hold in all contexts. This claim, i.e. that skeptical assumptions hold universally and timelessly, presupposes epistemological realism in conjunction with the claim that skeptical assumptions accurately reflect this objectively existing epistemic reality. So an adequate defense of it must also defend these presuppositions. So far however, the skeptic has merely assumed them to be true, and contextualists do not buy it. By rejecting these presuppositions, contextualism thereby removes the basis of the skeptical challenge.

The Default and Challenge alternative model for justification

Now that contextualism has shown Agrippan and Cartesian skepticism to rely on dubious and yet undefended assumptions, let us consider contextualism itself as an alternative account of epistemic justification which may defeat skepticism. Let us first consider the version put forth by Williams (2017). According to the Prior Grounding Requirement, a questioner is by default justified in demanding justification for beliefs, and the burden of justification by default lies on the believer. Williams proposes an alternative justification paradigm, the Default and Challenge model, which instead claims that the believer is per default justified in their beliefs, and that the burden of justification lies on the questioner, who must justify their demand for justification. The default burden of justification has thus been moved from the believer to the questioner. The Default and Challenge model is analogous to the 'innocent until proven guilty' legal principle, where the burden of proving guilt lies on the prosecutor. This model considers beliefs provisionally justified until sufficient grounds or evidence have been raised against their credibility. The believer is per default personally justified, even if further challenge may show them to be evidentially unjustified. In this sense, personal justification is prioritized over evidential justification. If a questioner succeeds in justifying their challenge, the Defense Commitment is activated, and the believer no longer enjoys default justification. They must now provide justification for their belief. It is by taking evidence against the credibity of one’s beliefs seriously, and by responding with justification for these beliefs, that a person fulfills their epistemic responsibility. So although evidential justification is subordinate to personal justification, the former is still an essential component of the latter, on this view.

The sort of justification that the questioner must offer, in order to justify their challenge, is proof of a risk of error, a risk sufficiently high to challenge the believer's default entitlement to being justified in their belief. The questioner can do this by showing the claim itself to be untrue, or by showing the claim to be unjustified, for example by showing that the procedure which led to the claim was epistemically unreliable. Williams' Default and Challenge model is contextualist in that what constitutes adequate justification of challenges (i.e. the demands for justification of beliefs) depends on the context_. Again, the variability can be both with regard to which criteria of justification are included, and with regard to the degree to which they must be satisfied. Such context-sensitive standards eliminate the possibility of justifying a general challenge. That is, it not possible on the contextualist view to justify once and for all a demand for all beliefs in all contexts. On the contextualist view, it is only possible to justify challenges to specific claims, by raising doubts specific to those claims. If a questioner succeeds in justifying their challenge of a given belief, and the cognizer cannot provide adequate justification for that belief, then indeed the conclusion is skeptical-- but only concerning that belief, in that context! Thus the Default and Challenge model prevents the Trilemma from demanding justification of all claims in general, and thereby also avoids its general scope of its skeptical conclusion.

Questioners do not always succeed in mustering the contextually adequate justification for their challenge. In such cases, the beliefs remain per default justified. Thus, justified beliefs exist. This refutes the Trilemma's skeptical conclusion, and this is how the Default and Challenge model manages to sidestep Agrippan skepticism. To the Cartesian challenge, of how one knows that one's interpretation of experience is a true description of the external world, the Default and Challenge model responds by posing a question right back: Does the questioner possess or can he provide sufficient reason to doubt the reliability of this interpretation? If not, the belief remains justified per default justified. This is how the Default and Challenge model deals with Cartesian skepticism.

4. The Assumption of Internalism

Internalism Assumption of Skeptical Arguments

There is another dimension along which skeptical assumptions may be questionable. In demanding the cognizer to provide the grounds and evidence for each and every one of their beliefs, the Trilemma is implicitly requiring that the cognizer be aware of this evidential justification. This is because ought implies can, which in turn implies awareness, since no one can provide grounds they are not aware of. Now since the Trilemma is committed to the Prior Grounding Requirement, it considers the evidential justification the basis of personal justification and justification simpliciter. Attaining personal justification and justification simpliciter requires being able to provide evidential justification. So in requiring cognizers be aware of the evidential justification, it is effectively requiring cognizers be aware of the basis of justification (simpliciter). "[N]o one’s belief can rest on grounds of which he is unaware".23 Cartesian skepticism shares the same requirement of awareness, just for other factors of justification. Cartesian skepticism argues we cannot have knowledge, unless we possess some valid way of discriminating between alternative hypotheses that fit equally well with our experiences. It is not enough that such a discriminating mechanisms exists-- we must be aware of it. If it merely exists somewhere, independently of us, as a reality or fact that we cannot access, then again we regress into the Cartesian question of "how do we know which of the hypotheses are true?"

The requirement, that the cognizer be aware of the basis of epistemic justification, is an internalist position, the variant called access internalism.24 There are many different internalist positions, which may differ in the degree to which the basis of justification is required to be accessible to the cognizer.25 Not all variants require that the cognizer be actually aware of basis of justification. Accessibility internalism does not require that the cognizer be actually aware, always or right now, of the basis of justification, but only that cognizers be capable of achieving such awareness eventually over time. Some variants require awareness or accessibility of only some of the factors that constitute the basis, for example that only parts of all the evidence be known. These are called partial account, in contrast to complete accounts, which require all factors constituting the basis be known or accessible to the cognizer. Some variants, so-called ’weak’ version, require only that the cognizer know OF the justifying factors, while others, i.e. ’strong’ versions, also require the cognizer know THAT such-and-such are indeed the justifying factors of a given belief. This is a sort of ’meta-awareness’ requirement. (Sosa & Stroud 1994, Stroud 2003) is an example of a debate on whether such meta-awareness is required for justification and knowledge.

There is also variance regarding the nature of the basis, of entities which could constitute the basis. Some internalists require the basis consist only of beliefs, while others, like the mentalism account, allow non-belief mental states like non-propositional experience, sensations etc. to form the basis as well. And a third dimension of variance is in how the concept of epistemic justification is understood. Deontological accounts conceive of epistemic justification more, if not solely, in normative terms. On these views, epistemic justification consists in fulfilling epistemic duties. Non-normative i.e. descriptive factors like the truth value of one's beliefs, or the truth-conduciveness of one's mechanisms for acquiring beliefs, do not feature in deontological criteria of justification much.

Externalism

The great diversity of views along many dimensions makes it hard to define Internalism as a whole group. But this can be done for Externalism, because all externalist positions share the characteristic that they are the negation of their corresponding internalist variant. For example, externalists about partial accessibility to justification negate the claim that in order to be justified, cognizers must be capable of gaining access to at least some of the factors constituting the basis of justification. In other words, these externalists claim that there exist cases where we cannot gain access to any of the justifying factors, and yet still be epistemically justified in holding those beliefs.

In the dialectic between Agrippan and Cartesian skepticism and Williams' contextualist account, the skeptical accounts require the cognizer be actually aware of (and provide) evidential justification for their beliefs. In adopting the Default and Challenge model, William's contextualist account is externalist because cognizers enjoying default justification need not be aware of the factors that make their beliefs evidentially justified. So they need not provide the Agrippan questioner with positive proof, nor need they provide the Cartesian with a criterion to discriminate between alternative hypotheses. "A belief is adequately grounded [i.e. evidentially justified] when it is formed by a method that is in fact reliable: responsibility demands that we be able to demonstrate reliability only if reasons emerge for suspecting unreliability."26 Here, the factor that justifies the belief is a fact that the cognizer need not be aware of (namely the fact that the method used to acquire the belief is in fact reliable), i.e. an external factor.

The account is even more externalist than this, in that it allows the subject to remain personally justified so long as they are epistemically responsible, even in cases where accidentality and bad epistemic luck make their beliefs false and hence evidentially unjustified. In making epistemic responsibility, i.e. the satisfaction of epistemic duties, the only requirement for epistemic justification, this account is a variant of deontological conceptions of epistemic justification. Again, this is the aforementioned prioritization of personal justification over evidential justification, which enables Williams' account to maintain that there are justified beliefs, and at least provisional knowledge. His account thus combines externalism with the prioritization of personal justification to avoid skepticism.

Arguments for Externalism

Many arguments for externalism are objections to internalism. Access internalism is refuted with examples where we intuitively feel the beliefs are justified despite a lack of actual awareness of their justification basis. The example often used is emergency situations, where people make rapid-fire judgments and decisions based on their instincts, and are justified in doing so. To slow down and examine justification bases in such situations would be life-threatening and irrational. Or at least, this is the intuition that is relied on in such arguments, and indeed, few philosophers have been willing to defend access internalism by appealing to the premise that such rapid-fire judgments and decisions are in fact unjustified.The emergency example is also used to argue against accessibility internalism. Even if a person were able to sit down and reflect after the emergency, it is not necessarily true that they would be able to gain access to the justifiers of the beliefs they acquired during the emergency. At the very least, complete internalism is ruled out: It is definitely implausible that they would be able to remember and successfully retrieve the justifiers for all the beliefs they acquired during that time. But it is also plausible that there be cases where the emergency was so acute and traumatic that the person not remember anything at all, and not be able to gain access to a single justifier. The existence of such cases would refute accessibility internalism.

Positive arguments for externalism exist as well however, that is, independent arguments which are not merely objections to internalism. They include positive defenses of accounts of justification (or knowledge) that presuppose or imply externalism. Reliable Process theories of epistemic justification are an example of such an account. Despite the many variants, the core tenet and common denominator of reliable process theories is that beliefs are justified only by the process(es) which produced or sustained them, if and only if it (they) are in fact reliable. How reliability is defined is one point of variance among the different reliable process theories, but an essential component of reliability is truth-conduciveness. A reliable mechanism is a mechanisms that truth-conducive in some way to some degree. An example of a positive argument for reliable process theories goes as follows: It is generally agreed that a couple of mechanisms for acquiring beliefs are validly considered reliable, and that beliefs that result from them validly enjoy prima facie justification. (Goldman 2009) Many of these center on sensory perception as the reliable mechanism. Goldman goes on to argue that the reason we take beliefs resulting from sensory perception to be prima facie justified is that they tend to be truth-conducive. In other words, we tacitly assume a connection between truth-conduciveness and epistemic justification. In inferring 'prima facie justified' from 'truth-conducive', we are assuming that the latter is sufficient for the former. This is at least our performative assumption. Perhaps if we make our own assumptions explicit, we will only endorse the more modest claim that truth-conduciveness is a necessary, and perhaps in some cases and to some extent a sufficient, condition for epistemic justification and knowledge. But nonetheless, we are committed to viewing reliable processes as justifiers of beliefs because of their truth-conduciveness. This is a positive argument for reliable process theories of epistemic justification, because invoking a deeply held intuition in favor of a given theory is a method of argument independent of objections to alternatives.

In turn, reliable process theories imply externalism, so their positive defense based on intuitions about truth-conducivness also acts as a positive defense of externalism. Whether this essential feature of truth-conduciveness is defined as a probability of truth in outcome, an actual performance record, or other is a point of variation which does not affect the fact that reliabilist theories as a group imply externalism. Reliable process theories imply externalism of accessibility justification, because a person may be justified in their beliefs so long as the mechanisms by which these beliefs were acquired and sustained are in fact reliable, even if this person does not possess or cannot even gain access to this fact (herein lies the negation of access/accessibility internalism). They imply externalism with regards to mentalism because again, the fact of these mechanisms being reliable is not a mental state, and this non-mental state is what is doing the justifying. On the mentalist view, only mental states can be justifying factors. They negate deontological conceptions of epistemic justification in that reliabilist justifiers, namely facts about the reliability of epistemic mechanisms, are descriptive entities independent of epistemic responsibility. Deontological conceptions define epistemic justification normatively, as fulfilling epistemic duties and responsibilities in some way. A person can be shirking such duties, and yet unintentionally or by accident still be applying epistemically reliable processes. The beliefs that result in such cases would be considered epistemically justified on reliabilist accounts, but unjustified on deontological accounts.

Objections to Externalism

Reliable process accounts of epistemic justification take truth-conduciveness to be the criterion by which to compare between alternative mechanisms for acquiring beliefs. It is in virtue of the fact that certain mechanisms lead to true data points that they are considered reliable. But this requires presupposing certain data point to be true in the first place. This means that the epistemic justification of reliable process theories depend on it truly being the case that the data-points they assume to be true, are true.

Stroud (2003) argues that the highest degree of certainty that an externalist cognizer can achieve is a conditional: "If what I believe about it is true and my beliefs about it are produced in what my theory says is the right way, I do know how human knowledge comes to be (...) But if my beliefs are not true, or not arrived at in that way, I do not. I wonder which it is. I wonder whether I understand human knowledge or not.’ That is not a satisfactory position to arrive at in one's study of human knowledge—or of anything else."27 This the strongest conclusion that any externalist theory can arrive at, because the epistemologist himself does not know whether the data points he takes for granted, or the knowledge acquisition mechanism he posits, are indeed true. In other words, it is from the first person perspective of the epistemological theorist that externalism fails. This degree of uncertainty must be dispelled by a reason for choosing one theory over another, a reason that the cognizer himself has to have access to and provide.

”But this is to presuppose internalism!” — Sosa objects. Sosa’s (1999) objection is simply to point out that in arguing thus, Stroud is assuming an internalist tenet, namely that the cognizer has to have possess i.e. have access to a reason (rational justification) for choosing his belief over alternatives. He is assuming that such an internal reason is a proper way and the only proper way to get out of the externalist’s predicament, namely the predicament that his theory is only justified if it is in fact true. But is this a predicament? Is it not simply an unavoidable, even essential, indispensable and defining feature of any epistemic theory that it require truth (of claims) or truth-conduciveness (of mechanisms) for justification and knowledge, and that the state of such features not obtaining must lead to lack of justification?28 The project of coming up with an account of what constitutes adequate justification or knowledge does not imply that we ourselves, as cognizing subjects, must necessarily possess an infallible mechanism for knowing when we satisfy the criterion proposed by the account and when we do not. Establishing such facts are after all the end goal of epistemic processes, not their precondition.

In summary, this essay has shown that skeptical arguments make certain assumptions that together indeed lead to skeptical conclusions. But various objections proposed by proponents of contextualism and externalism show that such assumptions can plausibly be rejected, thereby allowing us to reject the skeptical conclusion as well. The possibility of justification and knowledge has thus not been ruled out.

References

Alston, William P. (1991). Perceiving God: The Epistemology of Religious Experience. Ithaca, N.Y.: Cornell University Press.

Bonjour, Laurence (2006). Replies. Philosophical Studies 131 (3):743-759.

Cohen, Stewart (1999). Contextualism, skepticism, and the structure of reasons. Philosophical Perspectives 13:57-89.

Feyerabend, Paul (1999). Knowledge, Science and Relativism. New York: Cambridge University Press. Edited by John Preston.

Goldman, Alvin I. (2009). Internalism, Externalism, and the Architecture of Justification. Journal of Philosophy 106 (6): 309-338.

Henderson, Leah (Winter 2022 Edition). "The Problem of Induction". In The Stanford Encyclopedia of Philosophy. Edward N. Zalta & Uri Nodelman (eds.). https://plato.stanford.edu/archives/win2022/entries/induction-problem/.

Hume, David (1748). An Enquiry Concerning Human Understanding. Oxford: Oxford University Press.

Longino, Helen E. (1990). Science as Social Knowledge: Values and Objectivity in Scientific Inquiry. Princeton University Press.

Pappas, George (Spring 2023 Edition). "Internalist vs. Externalist Conceptions of Epistemic Justification". In The Stanford Encyclopedia of Philosophy. Edward N. Zalta & Uri Nodelman (eds.). https://plato.stanford.edu/archives/spr2023/entries/justep-intext/.

Sosa, Ernest & Stroud, Barry (1994). Philosophical Scepticism. Aristotelian Society Supplementary Volume 68:263-307.

Sosa, Ernest (1999). Skepticism and the internal/external divide. In John Greco & Ernest Sosa (eds.), The Blackwell Guide to Epistemology. Blackwell. pp. 145-57.

Stanford, Kyle (Summer 2023 edition). "Underdetermination of Scientific Theory". In The Stanford Encyclopedia of Philosophy. Edward N. Zalta & Uri Nodelman (eds.). https://plato.stanford.edu/archives/sum2023/entries/scientific-underdetermination/.

Stine, G. C. (1976). Skepticism, relevant alternatives, and deductive closure. Philosophical Studies 29 (4):249--261.

Stroud, Barry (2003). 'Understanding Human Knowledge in General'. In Understanding Human Knowledge: Philosophical Essays. Oxford online edn, Oxford Academic. https://doi.org/10.1093/0199252130.003.0008, accessed 25 Feb. 2024.

Williams, Michael (2017). Skepticism. In John Greco & Ernest Sosa (eds.), The Blackwell Guide to Epistemology. Oxford, UK: Blackwell. pp. 33–69.

  1. Williams 2017, 37↩︎

  2. Williams 2017, 40↩︎

  3. Note that the conclusion is that valid justification does not exist at all. The conclusion is not that valid justification may exist at lower epistemic standards, but not at the highest i.e. skeptical-level epistemic standards. The result of this is that it does not matter how demanding (or not) our requirements for knowledge are. Regardless of how demanding a given conception of knowledge is, the Trilemma undermines that conception of knowledge. This is because it undermines all justification itself, and as noted, we intuitively feel that knowledge requires justification, that justification is a necessary condition of knowledge. So, without justification in any form, there cannot be knowledge in any form.↩︎

  4. An expression used by BonJour 2006, 744↩︎

  5. This idea was echoed later on, in Philosophy of Science, as one of the main criticisms of logical empiricism. Propositions do not follow directly from empirical observations. True propositions are not served to us on a silver platter as a matter of course following experience. Experiences do not 'imply' certain propositions, because translation of raw experience to propositions is not at all a straightforward process. Such acts of translation are not assumptionless i.e. theory neutral. Cognizing subjects actively participate in the translation process. Perceptual beliefs are thus partially the result of subjective contribution, and not solely a 'reception of objective propositions'. A simple everyday example is the fact that various people at the same scene can interpret the events in completely different ways. A scientific example is that in recording observations, scientists cannot capture the complete sensory 'picture', and are forced to choose certain aspects to focus on, i.e. variables, and also how to measure and record them. All such choices of what to focus on, and also how to focus on them require theoretical assumptions, none of which were self-evident or implied by the observations themselves, and all of which could be questioned and replaced with alternatives that might equally fit with the observations. All this makes observation (and science) more like an art rather than a science, in the way science is commonly understood. Feyerabend (1999), Longino (1990) and the Duhem-Quine thesis discussed in Stanford (2023) are examples of discussions about the theory-ladenness of observation.↩︎

  6. It is important to note that according to Descartes, his examples do not change our experiences— these are as they are no matter our beliefs. But our explanations i.e. interpretations of these experiences are brought into question by his thought experiments.↩︎

  7. For example the attempt of some religious groups to reconcile the Bible with the existence of fossils, by claiming that God planted fossils on Earth at the time of Genesis.↩︎

  8. This does not mean that Hume necessarily objected to induction in practice. There may be pragmatic justification for it. He simply notes that there is no rational justification for it. See Henderson 2022 and Hume 1748.↩︎

  9. Alston 1991, 148. Alston is here talking of mechanisms for acquiring knowledge instead of systems of belief. But the point against coherence as a sufficient condition remains the same.

    ↩︎

  10. Williams 2017, 42↩︎

  11. Williams 2017, 42↩︎

  12. Again, see footnote 4 on later echoes in Philosophy of Science↩︎

  13. Sosa & Stroud 1994, 269, emphasis added↩︎

  14. Sosa & Stroud 1994, 271↩︎

  15. This second premise, together with the conditional, would imply the conclusion that one would be right whenever one thought. Combining the two premises would yield the following argument: P1: If I think that I am thinking, then I must be right; P2: It is impossible to think that I am not thinking. C: If I think, I must be right.↩︎

  16. Sosa & Stroud 1994, 272, emphasis added↩︎

  17. William 2017, 47↩︎

  18. Williams 2017, 48↩︎

  19. Williams 2017, 50↩︎

  20. Williams 2017, 51↩︎

  21. Williams 2017, 61↩︎

  22. Cohen (1999, 85, note 28) on this point being made by Stine (1976).↩︎

  23. Williams 2017, 51↩︎

  24. Pappas 2023↩︎

  25. The basis in question can also be a basis for knowledge. But since the dimension distinguishing between justification and knowledge is orthogonal to the distinction between internalism and externalism, and justification is seen as an essential component of knowledge, it is not relevant to our present discussion. Let us thus set aside the distinction between justification and knowledge, and focus on justification.↩︎

  26. Williams 2017, 54↩︎

  27. Stroud 2003, 119↩︎

  28. ”The essential feature of epistemic justification is that it be truth-conducive, to justify a belief is to link it with some feature that increases the likelihood of its being true. An account of justification that does not make this aspect absolutely fundamental is therefore either not an account of epistemic justification at all or it is an account that makes epistemic justification purely dialectical.” (Williams 2017, 54)↩︎